how to use google for research papers

Reference management. Clean and simple.

Google Scholar: the ultimate guide

What is Google Scholar?

Why is google scholar better than google for finding research papers, the google scholar search results page, the first two lines: core bibliographic information, quick full text-access options, "cited by" count and other useful links, tips for searching google scholar, 1. google scholar searches are not case sensitive, 2. use keywords instead of full sentences, 3. use quotes to search for an exact match, 3. add the year to the search phrase to get articles published in a particular year, 4. use the side bar controls to adjust your search result, 5. use boolean operator to better control your searches, google scholar advanced search interface, customizing search preferences and options, using the "my library" feature in google scholar, the scope and limitations of google scholar, alternatives to google scholar, country-specific google scholar sites, frequently asked questions about google scholar, related articles.

Google Scholar (GS) is a free academic search engine that can be thought of as the academic version of Google. Rather than searching all of the indexed information on the web, it searches repositories of:

• universities
• scholarly websites

This is generally a smaller subset of the pool that Google searches. It's all done automatically, but most of the search results tend to be reliable scholarly sources.

However, Google is typically less careful about what it includes in search results than more curated, subscription-based academic databases like Scopus and Web of Science . As a result, it is important to take some time to assess the credibility of the resources linked through Google Scholar.

➡️ Take a look at our guide on the best academic databases .

One advantage of using Google Scholar is that the interface is comforting and familiar to anyone who uses Google. This lowers the learning curve of finding scholarly information .

There are a number of useful differences from a regular Google search. Google Scholar allows you to:

• copy a formatted citation in different styles including MLA and APA
• export bibliographic data (BibTeX, RIS) to use with reference management software
• explore other works have cited the listed work
• easily find full text versions of the article

Although it is free to search in Google Scholar, most of the content is not freely available. Google does its best to find copies of restricted articles in public repositories. If you are at an academic or research institution, you can also set up a library connection that allows you to see items that are available through your institution.

The Google Scholar results page differs from the Google results page in a few key ways. The search result page is, however, different and it is worth being familiar with the different pieces of information that are shown. Let's have a look at the results for the search term "machine learning.”

• The first line of each result provides the title of the document (e.g. of an article, book, chapter, or report).
• The second line provides the bibliographic information about the document, in order: the author(s), the journal or book it appears in, the year of publication, and the publisher.

Clicking on the title link will bring you to the publisher’s page where you may be able to access more information about the document. This includes the abstract and options to download the PDF.

To the far right of the entry are more direct options for obtaining the full text of the document. In this example, Google has also located a publicly available PDF of the document hosted at umich.edu . Note, that it's not guaranteed that it is the version of the article that was finally published in the journal.

Below the text snippet/abstract you can find a number of useful links.

• Cited by : the cited by link will show other articles that have cited this resource. That is a super useful feature that can help you in many ways. First, it is a good way to track the more recent research that has referenced this article, and second the fact that other researches cited this document lends greater credibility to it. But be aware that there is a lag in publication type. Therefore, an article published in 2017 will not have an extensive number of cited by results. It takes a minimum of 6 months for most articles to get published, so even if an article was using the source, the more recent article has not been published yet.
• Versions : this link will display other versions of the article or other databases where the article may be found, some of which may offer free access to the article.
• Quotation mark icon : this will display a popup with commonly used citation formats such as MLA, APA, Chicago, Harvard, and Vancouver that may be copied and pasted. Note, however, that the Google Scholar citation data is sometimes incomplete and so it is often a good idea to check this data at the source. The "cite" popup also includes links for exporting the citation data as BibTeX or RIS files that any major reference manager can import.

Pro tip: Use a reference manager like Paperpile to keep track of all your sources. Paperpile integrates with Google Scholar and many popular academic research engines and databases, so you can save references and PDFs directly to your library using the Paperpile buttons and later cite them in thousands of citation styles:

Although Google Scholar limits each search to a maximum of 1,000 results , it's still too much to explore, and you need an effective way of locating the relevant articles. Here’s a list of pro tips that will help you save time and search more effectively.

You don’t need to worry about case sensitivity when you’re using Google scholar. In other words, a search for "Machine Learning" will produce the same results as a search for "machine learning.”

Let's say your research topic is about self driving cars. For a regular Google search we might enter something like " what is the current state of the technology used for self driving cars ". In Google Scholar, you will see less than ideal results for this query .

The trick is to build a list of keywords and perform searches for them like self-driving cars, autonomous vehicles, or driverless cars. Google Scholar will assist you on that: if you start typing in the search field you will see related queries suggested by Scholar!

If you put your search phrase into quotes you can search for exact matches of that phrase in the title and the body text of the document. Without quotes, Google Scholar will treat each word separately.

This means that if you search national parks , the words will not necessarily appear together. Grouped words and exact phrases should be enclosed in quotation marks.

A search using “self-driving cars 2015,” for example, will return articles or books published in 2015.

Using the options in the left hand panel you can further restrict the search results by limiting the years covered by the search, the inclusion or exclude of patents, and you can sort the results by relevance or by date.

Searches are not case sensitive, however, there are a number of Boolean operators you can use to control the search and these must be capitalized.

• AND requires both of the words or phrases on either side to be somewhere in the record.
• NOT can be placed in front of a word or phrases to exclude results which include them.
• OR will give equal weight to results which match just one of the words or phrases on either side.

➡️ Read more about how to efficiently search online databases for academic research .

In case you got overwhelmed by the above options, here’s some illustrative examples:

Tip: Use the advanced search features in Google Scholar to narrow down your search results.

You can gain even more fine-grained control over your search by using the advanced search feature. This feature is available by clicking on the hamburger menu in the upper left and selecting the "Advanced search" menu item.

Adjusting the Google Scholar settings is not necessary for getting good results, but offers some additional customization, including the ability to enable the above-mentioned library integrations.

The settings menu is found in the hamburger menu located in the top left of the Google Scholar page. The settings are divided into five sections:

• Collections to search: by default Google scholar searches articles and includes patents, but this default can be changed if you are not interested in patents or if you wish to search case law instead.
• Bibliographic manager: you can export relevant citation data via the “Bibliography manager” subsection.
• Languages: if you wish for results to return only articles written in a specific subset of languages, you can define that here.
• Library links: as noted, Google Scholar allows you to get the Full Text of articles through your institution’s subscriptions, where available. Search for, and add, your institution here to have the relevant link included in your search results.
• Button: the Scholar Button is a Chrome extension which adds a dropdown search box to your toolbar. This allows you to search Google Scholar from any website. Moreover, if you have any text selected on the page and then click the button it will display results from a search on those words when clicked.

When signed in, Google Scholar adds some simple tools for keeping track of and organizing the articles you find. These can be useful if you are not using a full academic reference manager.

All the search results include a “save” button at the end of the bottom row of links, clicking this will add it to your "My Library".

To help you provide some structure, you can create and apply labels to the items in your library. Appended labels will appear at the end of the article titles. For example, the following article has been assigned a “RNA” label:

Within your Google Scholar library, you can also edit the metadata associated with titles. This will often be necessary as Google Scholar citation data is often faulty.

There is no official statement about how big the Scholar search index is, but unofficial estimates are in the range of about 160 million , and it is supposed to continue to grow by several million each year.

Yet, Google Scholar does not return all resources that you may get in search at you local library catalog. For example, a library database could return podcasts, videos, articles, statistics, or special collections. For now, Google Scholar has only the following publication types:

• Journal articles : articles published in journals. It's a mixture of articles from peer reviewed journals, predatory journals and pre-print archives.
• Books : links to the Google limited version of the text, when possible.
• Book chapters : chapters within a book, sometimes they are also electronically available.
• Book reviews : reviews of books, but it is not always apparent that it is a review from the search result.
• Conference proceedings : papers written as part of a conference, typically used as part of presentation at the conference.
• Court opinions .
• Patents : Google Scholar only searches patents if the option is selected in the search settings described above.

The information in Google Scholar is not cataloged by professionals. The quality of the metadata will depend heavily on the source that Google Scholar is pulling the information from. This is a much different process to how information is collected and indexed in scholarly databases such as Scopus or Web of Science .

➡️ Visit our list of the best academic databases .

Google Scholar is by far the most frequently used academic search engine , but it is not the only one. Other academic search engines include:

• Science.gov
• Semantic Scholar
• scholar.google.fr : Sur les épaules d'un géant
• scholar.google.es (Google Académico): A hombros de gigantes
• scholar.google.pt (Google Académico): Sobre os ombros de gigantes
• scholar.google.de : Auf den Schultern von Riesen

➡️ Once you’ve found some research, it’s time to read it. Take a look at our guide on how to read a scientific paper .

No. Google Scholar is a bibliographic search engine rather than a bibliographic database. In order to qualify as a database Google Scholar would need to have stable identifiers for its records.

No. Google Scholar is an academic search engine, but the records found in Google Scholar are scholarly sources.

No. Google Scholar collects research papers from all over the web, including grey literature and non-peer reviewed papers and reports.

Google Scholar does not provide any full text content itself, but links to the full text article on the publisher page, which can either be open access or paywalled content. Google Scholar tries to provide links to free versions, when possible.

The easiest way to access Google scholar is by using The Google Scholar Button. This is a browser extension that allows you easily access Google Scholar from any web page. You can install it from the Chrome Webstore .

18 Google Scholar tips all students should know

Dec 13, 2022

[[read-time]] min read

Think of this guide as your personal research assistant.

“It’s hard to pick your favorite kid,” Anurag Acharya says when I ask him to talk about a favorite Google Scholar feature he’s worked on. “I work on product, engineering, operations, partnerships,” he says. He’s been doing it for 18 years, which as of this month, happens to be how long Google Scholar has been around.

Google Scholar is also one of Google’s longest-running services. The comprehensive database of research papers, legal cases and other scholarly publications was the fourth Search service Google launched, Anurag says. In honor of this very important tool’s 18th anniversary, I asked Anurag to share 18 things you can do in Google Scholar that you might have missed.

1. Copy article citations in the style of your choice.

With a simple click of the cite button (which sits below an article entry), Google Scholar will give you a ready-to-use citation for the article in five styles, including APA, MLA and Chicago. You can select and copy the one you prefer.

2. Dig deeper with related searches.

Google Scholar’s related searches can help you pinpoint your research; you’ll see them show up on a page in between article results. Anurag describes it like this: You start with a big topic — like “cancer” — and follow up with a related search like “lung cancer” or “colon cancer” to explore specific kinds of cancer.

Related searches can help you find what you’re looking for.

3. And don’t miss the related articles.

This is another great way to find more papers similar to one you found helpful — you can find this link right below an entry.

4. Read the papers you find.

Scholarly articles have long been available only by subscription. To keep you from having to log in every time you see a paper you’re interested in, Scholar works with libraries and publishers worldwide to integrate their subscriptions directly into its search results. Look for a link marked [PDF] or [HTML]. This also includes preprints and other free-to-read versions of papers.

5. Access Google Scholar tools from anywhere on the web with the Scholar Button browser extension.

The Scholar Button browser extension is sort of like a mini version of Scholar that can move around the web with you. If you’re searching for something, hitting the extension icon will show you studies about that topic, and if you’re reading a study, you can hit that same button to find a version you read, create a citation or to save it to your Scholar library.

Install the Scholar Button Chrome browser extension to access Google Scholar from anywhere on the web.

6. Learn more about authors through Scholar profiles.

There are many times when you’ll want to know more about the researchers behind the ideas you’re looking into. You can do this by clicking on an author’s name when it’s hyperlinked in a search result. You’ll find all of their work as well as co-authors, articles they’re cited in and so on. You can also follow authors from their Scholar profile to get email updates about their work, or about when and where their work is cited.

7. Easily find topic experts.

One last thing about author profiles: If there are topics listed below an author’s name on their profile, you can click on these areas of expertise and you’ll see a page of more authors who are researching and publishing on these topics, too.

8. Search for court opinions with the “Case law” button.

Scholar is the largest free database of U.S. court opinions. When you search for something using Google Scholar, you can select the “Case law” button below the search box to see legal cases your keywords are referenced in. You can read the opinions and a summary of what they established.

9. See how those court opinions have been cited.

If you want to better understand the impact of a particular piece of case law, you can select “How Cited,” which is below an entry, to see how and where the document has been cited. For example, here is the How Cited page for Marbury v. Madison , a landmark U.S. Supreme Court ruling that established that courts can strike down unconstitutional laws or statutes.

10. Understand how a legal opinion depends on another.

When you’re looking at how case laws are cited within Google Scholar, click on “Cited by” and check out the horizontal bars next to the different results. They indicate how relevant the cited opinion is in the court decision it’s cited within. You will see zero, one, two or three bars before each result. Those bars indicate the extent to which the new opinion depends on and refers to the cited case.

In the Cited by page for New York Times Company v. Sullivan, court cases with three bars next to their name heavily reference the original case. One bar indicates less reliance.

11. Sign up for Google Scholar alerts.

Want to stay up to date on a specific topic? Create an alert for a Google Scholar search for your topics and you’ll get email updates similar to Google Search alerts. Another way to keep up with research in your area is to follow new articles by leading researchers. Go to their profiles and click “Follow.” If you’re a junior grad student, you may consider following articles related to your advisor’s research topics, for instance.

12. Save interesting articles to your library.

It’s easy to go down fascinating rabbit hole after rabbit hole in Google Scholar. Don’t lose track of your research and use the save option that pops up under search results so articles will be in your library for later reading.

13. Keep your library organized with labels.

Labels aren’t only for Gmail! You can create labels within your Google Scholar library so you can keep your research organized. Click on “My library,” and then the “Manage labels…” option to create a new label.

14. If you’re a researcher, share your research with all your colleagues.

Many research funding agencies around the world now mandate that funded articles should become publicly free to read within a year of publication — or sooner. Scholar profiles list such articles to help researchers keep track of them and open up access to ones that are still locked down. That means you can immediately see what is currently available from researchers you’re interested in and how many of their papers will soon be publicly free to read.

15. Look through Scholar’s annual top publications and papers.

Every year, Google Scholar releases the top publications based on the most-cited papers. That list (available in 11 languages) will also take you to each publication’s top papers — this takes into account the “h index,” which measures how much impact an article has had. It’s an excellent place to start a research journey as well as get an idea about the ideas and discoveries researchers are currently focused on.

16. Get even more specific with Advanced Search.

Click on the hamburger icon on the upper left-hand corner and select Advanced Search to fine-tune your queries. For example, articles with exact words or a particular phrase in the title or articles from a particular journal and so on.

17. Find extra help on Google Scholar’s help page.

It might sound obvious, but there’s a wealth of useful information to be found here — like how often the database is updated, tips on formatting searches and how you can use your library subscriptions when you’re off-campus (looking at you, college students!). Oh, and you’ll even learn the origin of that quote on Google Scholar’s home page.

18. Keep up with Google Scholar news.

Don’t forget to check out the Google Scholar blog for updates on new features and tips for using this tool even better.

Related stories

Quiz: Do you know solar eclipse Search Trends?

4 ways to use Search to check facts, images and sources online

Get more personalized shopping options with these Google tools

6 ways to travel smarter this summer using Google tools

Empowering your team to build best-in-class mmms.

New ways we’re tackling spammy, low-quality content on Search

Let’s stay in touch. Get the latest news from Google in your inbox.

How to Use Google Scholar for Research: A Complete Guide

To remain competitive, Research and Development (R&D) teams must utilize all of the resources available to them. Google Scholar can be a powerful asset for R&D professionals who are looking to quickly find relevant sources related to their project.  With its sophisticated search engine capabilities, advanced filtering options, and alert notifications, using Google Scholar for research allows teams to easily locate reliable information in an efficient manner. Want to learn how to use google scholar for research? This blog post will cover how to use google scholar for research, how R&D professionals can exploit the potential of Google Scholar to uncover novel discoveries related to their projects, as well as remain apprised of advancements in their area.

Table of Contents

What is Google Scholar?

Overview of google scholar, searching with google scholar, finding relevant sources with google scholar, exploring related topics, evaluating sources found on google scholar, staying up to date with google scholar alerts, faqs in relation to how to use google scholar for research, how do i use google scholar for research, can you use google scholar for research papers, why is it important to use google scholar for research, are google scholar articles credible.

Google Scholar is a powerful research platform that enables users to quickly find, access, and evaluate scholarly information. It provides easy access to academic literature from all disciplines, including books, journal articles, conference papers, and more. Google Scholar offers researchers a wide range of tools for searching the web for the relevant content as well as ways to keep up with new developments in their field.

Google Scholar i s an online search engine designed specifically for finding scholarly literature on the internet. Google Scholar provides access to a vast array of scholarly literature from renowned universities and publishers around the world, simplifying the process of locating relevant material on any subject. In addition to its comprehensive indexing capabilities, Google Scholar also includes advanced search features such as citation tracking and alert notifications when new results are published in your chosen areas of interest.

The platform makes it a breeze for users to traverse multiple facets of a given topic by providing them with an array of different filters they can apply when conducting searches – these include things such as author name or publication date range; language; type (e.g., book chapter vs journal article); source material (e.g., open access only); etc Moreover, many results found through this platform come equipped with full-text PDFs available for download – so you don’t have to worry about pesky paywalls blocking your path while doing research.

Google Scholar is an invaluable resource for research and development teams, offering quick access to a wealth of scholarly information. Utilizing the proper search approaches, you can quickly locate precisely what you need by employing Google Scholar. Let’s look now at how to refine your results with advanced search techniques.

Key Takeaway:  Google Scholar is a powerful research platform that gives researchers an array of tools to quickly locate, access and evaluate scholarly information. It provides users with advanced search features such as citation tracking and alert notifications, along with easy-to-apply filters for narrowing down results by author name or publication date range – making it the go-to tool for any researcher looking to cut through the noise.

Exploring with Google Scholar can be a useful approach to quickly locate applicable scholarly material. There are several different strategies that can be used to get the most out of this powerful tool.

Basic google scholar search strategies involve entering a few keywords or phrases into the search bar and then refining your results using filters, sorting options, and related topics. This method is ideal for those who require a rapid search of information without needing to expend an excessive amount of time researching exact terms, especially for those unfamiliar with searching databases such as Google Scholar. It’s also useful for those who don’t have a lot of experience in searching databases like Google Scholar. 

Advanced search strategies allow users to take advantage of more sophisticated features such as Boolean operators , wildcards, and phrase searches. These tools make it easier to narrow down results by specifying exactly what you’re looking for or excluding irrelevant sources from your search results. Advanced searchers should also pay attention to synonyms when crafting their queries since these can help broaden the scope of their searches while still providing relevant results.

Finally, refining your results is key in order to ensure that you only see sources that are truly relevant and authoritative on the topic at hand. Filters such as date range, publication type, language, author name, etc., can help refine your query so that only high-quality sources appear in your list of results. Sorting options provide users with the ability to prioritize documents, enabling them to quickly locate relevant materials without needing to review a large number of irrelevant ones. 

Utilizing Google Scholar can be advantageous for swiftly finding pertinent research materials, but it is essential to comprehend the search strategies and filters at hand in order to maximize your searches. By understanding how to identify keywords and phrases, explore related topics, and utilize sorting options and filters, you can ensure that you are finding all of the relevant sources for your research project. 

Key Takeaway:  Google Scholar is a great tool for quickly locating relevant research sources. Advanced searchers can make use of Boolean operators, wildcards and phrase searches to narrow down their results while basic search strategies such as entering keywords into the search bar work just fine too. Additionally, refining your results with filters and sorting options helps ensure that you only see high-quality sources related to your topic at hand.

Locating applicable materials via Google Scholar can be a challenging endeavor, particularly for those unfamiliar with the research process. To facilitate the research process, employing various strategies can expedite and refine the search for relevant sources through Google Scholar. 

Making use of keywords and phrases is a powerful method for finding pertinent sources on Google Scholar. It is important to identify key terms related to your topic or research question so you can narrow down the results. Additionally, using quotation marks around multiple words will allow you to get more precise results as it searches for exact matches instead of individual words within a phrase.

Exploring related topics helps provide additional context when researching on Google Scholar. This includes looking at previous studies conducted on similar topics or areas of interest, which provides further insight into potential sources available from other researchers’ work in the field. Utilizing tools such as co-citation analysis also allows users to explore how different authors have been cited together over time by providing visualizations based on their connections and relationships with each other through citations.

Utilizing filters and sorting options such as language, date range, publication type, etc., enables users to refine their search even further so they only receive results that match their specific criteria. Sorting options like relevance ranking or date published also make it easier for them to find what they need without having to sift through hundreds of irrelevant documents manually. By utilizing these features effectively, researchers can save valuable time when searching for relevant sources in Google Scholar since all the information they need will already be organized accordingly right away, saving them an hour’s worth of manual labor.

By utilizing Google Scholar, research teams can quickly and easily find relevant sources for their projects. With the next heading, we will explore how to evaluate these sources for credibility and authority.

Key Takeaway:  Utilizing the right keywords and phrases, exploring related topics, and utilizing filters are essential techniques for finding relevant sources quickly with Google Scholar. By taking advantage of the available features, you can swiftly and accurately pinpoint documents that meet your criteria.

To assess the reliability and authority of each source, consider factors such as the publication’s reputation, author credentials in the field, and when it was published. To do this, look for publications from reputable journals or authors with credentials in the field. Furthermore, consider when the source was issued – more modern pieces may be more pertinent and exact than older ones.

It is advantageous to be aware of the distinct kinds of publications that can appear in search results, such as scholarly articles, books, conference papers, and dissertations; each offering various degrees of precision and accuracy depending on their intent and target audience. 

For example, a book chapter may provide an overview of a topic while a peer-reviewed journal article will contain more detailed information backed up by research evidence. Similarly, conference papers are typically shorter summaries of research projects whereas dissertations offer comprehensive coverage including methodology and analysis results. Understanding these differences helps you identify which sources are most suitable for your needs when conducting research using Google Scholar.

Evaluating sources found on Google Scholar is an important step to ensure the credibility and accuracy of research results. By setting up alerts with Google Scholar, you can stay informed about new research findings and manage your subscriptions accordingly.

Maximize your research efforts with Google Scholar. Assess credibility & authority, pay attention to the date of publication & understand different types of publications. #ResearchTips #GoogleScholar Click to Tweet

Google Scholar is an invaluable tool for staying up to date with the latest research in your field. With its alert feature, you can easily set up notifications so that you’re always on top of new developments. Setting up alerts and managing them effectively will help ensure that you never miss a beat when it comes to relevant information.

Begin your research by utilizing Google Scholar’s sophisticated search features such as keyword and phrase searches, sorting results according to relevance or date of publication, and excluding unrelated sources. Once you’ve identified the most pertinent topics related to your research interests, set up alerts for each one by clicking on the bell icon in the upper right corner of the page. This will allow Google Scholar to send notifications whenever new content is published about those specific topics.

When setting up alerts in Google Scholar, make sure that they are tailored specifically toward what matters most to you – this could include certain authors or journals whose work has particular relevance to your own research projects. You can also adjust how often these alerts are sent (daily or weekly) depending on how frequently new material is being published within those fields of study. Additionally, if there are any other sources outside of Google Scholar which may contain useful information (such as blogs), consider adding their RSS feeds into your alert system too so that all relevant updates appear in one place.

Finally, don’t forget to manage existing alerts regularly; this means keeping track of which ones are still relevant and deleting any no longer needed from time to time (this helps keep clutter down). Additionally, try experimenting with different combinations/filters within each alert until you find what works best for keeping yourself informed without getting overwhelmed with notifications.

Key Takeaway:  Utilize Google Scholar to stay up-to-date on the latest research in your field – create tailored alerts for specific topics and authors, adjust frequency of notifications as needed, and manage existing alerts regularly. Stay ahead of the curve by gathering all pertinent news in one location.

Google Scholar is a great tool for conducting research. It provides access to millions of scholarly articles, books, and other sources from across the web. Google scholar works by entering keywords related to your topic into the search bar at the top of the page to quickly locate relevant scholarly articles, books, and other sources from across the web. Then narrow down your results using filters such as date range or publication type.

Finally, skim through the abstracts and full texts to pinpoint useful information for your research project.

Yes, Google Scholar is a great resource for research papers. It offers access to an extensive range of scholarly literature from journals, books, and conference proceedings. The search engine provides a convenient way to locate the most recent research in any area by entering keywords or phrases.

Advanced capabilities, such as citation monitoring, can be utilized to track the latest citations of one’s own or others’ work.

Google Scholar is an invaluable tool for research, as it provides access to a vast range of scholarly literature from around the world. It allows researchers to quickly and easily search through millions of publications and journals in order to find relevant information.

Google Scholar also offers the ability to trace connections between different works, allowing researchers to stay abreast of recent developments in their field. With its user-friendly interface, Google Scholar makes researching easier than ever before.

Yes, Google Scholar articles are credible. They provide access to a wide range of academic literature from reliable sources such as peer-reviewed journals and conference proceedings. Expert scrutiny has been conducted to guarantee the accuracy and excellence of the articles before they are put up on Google Scholar. Additionally, each article includes information about its authorship and citation count which can help readers assess their credibility further.

Google Scholar provides a convenient way to uncover pertinent material, assess the quality of these sources with ease, and be informed about novel advancements in your area through notifications.  Thus, R&D supervisors should know how to use google scholar for research. Also, R&D supervisors considering utilizing Google Scholar for investigation ought to recall that this apparatus should not supplant customary techniques, for example, peer survey or manual searching; rather it should supplement them.

With its powerful search capabilities and ability to keep researchers informed about their fields of interest, using Google Scholar for research can save time while providing more accurate results than ever before.

Unlock the power of research with Cypris . Our platform provides rapid time to insights, enabling R&D and innovation teams to quickly access data sources for their projects.

Similar insights you might enjoy

Gallium Nitride Innovation Pulse

Carbon Capture & Storage Innovation Pulse

Sodium-Ion Batteries Innovation Pulse

What is Google Scholar? How to use the academic database for research

• Google Scholar is a searchable database of scholarly literature.
• It connects users with studies and journal articles on nearly any topic of interest.
• Not all articles are free — you might need a membership to read the full versions.

Established in 2004, Google Scholar is a massive database of scholarly literature that allows users to access information, cross reference it with other sources, and keep up with new research as it comes out.

Using Google Scholar, you can access these kinds of sources:

• Conference papers
• Academic books
• Theses and dissertations
• Technical reports

Here's everything you need to know about the powerful research tool.

How to use Google Scholar

Anyone can access the search database. And while it's built with college or grad students and other academics in mind — to help those writing academic papers create bibliographies more easily — anyone can reap its benefits.

Here are just a few examples of what you can do through Google Scholar: 

• Create alerts. You can create a library of research around a topic of interest, like global warming, and create alerts for it so that you're always up-to-date on the latest research.
• Explore related works. You can gain deeper knowledge around a complicated topic that you're interested in, like studies in the field of astronomy, by exploring related citations, authors, and publications.
• Check out the References section. Accessing an article's References section can help you branch out your research to see what sources an author used for their paper. 
• Save articles to your library. Saving your searches to your Google Scholar library helps you organize and keep track of your favorite results. 
• Citation export. You can export an article's full citation in your preferred format using the "Bibliography Manager" section. 

Accessing information 

Google Scholar is free to use as a search tool. However, since it pulls information from many other databases, it's possible that some of the results you pull up will require a login (or even payment) to access the full information.

Still, descriptions or abstracts are typically free and provide an overview of what's contained within the article. 

Overall, Google Scholar provides an excellent avenue into scholarly research, and while it does have its drawbacks, it's a tool that can be used to help clarify, explore and inform users about a wide variety of topics.

On February 28, Axel Springer, Business Insider's parent company, joined 31 other media groups and filed a $2.3 billion suit against Google in Dutch court, alleging losses suffered due to the company's advertising practices.

• Main content

• CQUniversity Library
• Library Guides
• How do I ...?

Using Google Scholar

• Tips on using Google Scholar effectively

Finding full text at CQU using Google Scholar

Using the view @ cquniversity link to access articles, accessing full text from your search results without view @ cqu links, video: an overview of searching and setting up library links, more tips....

GoogleScholar is a web search engine created by Google to locate academic and scholarly sources.

Sources include peer-reviewed papers, theses, books, book chapters, abstracts and articles from academic publishers, professional societies, preprint repositories, universities and other scholarly organisations.

Note: Only some of the Library's resources appear in Google Scholar search results , however, it can be viewed as a complementary search tool for finding scholarly information.

You can set up Google Scholar to identify whether the resources it finds are held in full text at CQUniversity Library and/or any other libraries to which you have access.

To do this:

• Click on Settings at the top of the page (the wheel icon)
• Click on Library Links in the left hand menu
• Type Central Queensland University into the search box
• Tick the checkbox beside CQUniversity - View @ CQUniversity

Now when you search Google Scholar, View @ CQUniversity will appear to the right of the results if CQUniversity Library has access to the full text.

A print version of these instructions: Setting up View@CQUniversity in Google Scholar (PDF) .

• Click the View @ CQUniversity link to the right of the result you want to read
• You will usually be taken to the database record. If you are taken to the Library Search record, use one of the Full Text options.

• Log in to see the full text.

You won’t be able to access everything in a GoogleScholar results list. And View @ CQUniversity doesn’t link to eBooks in our databases.

If there’s no View @ CQUniversity link to the right, there are a few things you can try:

• Click the [PDF] link to the right.
• Click the title in the results list.
• Search for books and eBooks by title in Library Search

Research Higher Degree students and staff may be able to get items via Document Delivery if they can't access them via GoogleScholar.

This enables you to use a variety of search tools, including boolean searching. You can also search for particular authors, and limit to specific types of publications, dates, and subject areas.

• Open the Hamburger menu in the top left (It looks like 3 horizontal stripes)
• Click Advanced search to open the Advanced Search window.

Using the advanced search boxes for your Keywords / search terms

• with all of the words - This is like a basic search. It puts AND between the entries.
• with the exact phrase - This is like using the double quotation marks for a phrase. Use this for one phrase. If you have more than one search term that is a phrase, use the double quotation marks around the words and put them in the "with all of the words" box
• with at least one of the words - This is like using OR. Use this search box for synonyms
• without the words - This is like using NOT to exclude terms you don't want.

Using the limiters

• where my words occur - This allows you to search the text of the whole article for your keywords/terms, or to limit the search to the title field of records.
• Return articles authored by - Use this to search for known authors by name
• Return articles published in - Use this to search for known journals by name
• Return articles dated between - This allows you to limit by publication year range.
• Next: Need Help? >>
• Last Updated: Dec 5, 2023 4:59 PM
• URL: https://libguides.library.cqu.edu.au/gscholar

Faculty and researchers : We want to hear from you! We are launching a survey to learn more about your library collection needs for teaching, learning, and research. If you would like to participate, please complete the survey by May 17, 2024. Thank you for your participation!

• University of Massachusetts Lowell
• University Libraries

Google Scholar Search Strategies

• About Google Scholar
• Manage Settings
• Enable My Library
• Google Scholar Library
• Cite from Google Scholar
• Tracking Citations
• Add Articles Manually
• Refine your Profile Settings

Using Google Scholar for Research

Google Scholar is a powerful tool for researchers and students alike to access peer-reviewed papers. With Scholar, you are able to not only search for an article, author or journal of interest, you can also save and organize these articles, create email alerts, export citations and more. Below you will find some basic search tips that will prove useful.

This page also includes information on Google Scholar Library - a resource that allows you to save, organize and manage citations - as well as information on citing a paper on Google Scholar.

Search Tips

• Locate Full Text
• Sort by Date
• Related Articles
• Court Opinions
• Email Alerts
• Advanced Search

Abstracts are freely available for most of the articles and UMass Lowell holds many subscriptions to journals and online resources. The first step is make sure you are affiliated with the UML Library on and off campus by Managing your Settings, under Library Links. 

When searching in Google Scholar here are a few things to try to get full text:

• click a library link, e.g., "Full-text @ UML Library", to the right of the search result;
• click a link labeled [PDF] to the right of the search result;
• click "All versions" under the search result and check out the alternative sources;
• click "More" under the search result to see if there's an option for full-text;
• click "Related articles" or "Cited by" under the search result to explore similar articles.

Your search results are normally sorted by relevance, not by date. To find newer articles, try the following options in the left sidebar:

• click "Sort by date" to show just the new additions, sorted by date;  If you use this feature a lot, you may also find it useful to setup email alerts to have new results automatically sent to you.
• click the envelope icon to have new results periodically delivered by email.

Note: On smaller screens that don't show the sidebar, these options are available in the dropdown menu labeled "Any time" right below the search button .

The Related Articles option under the search result can be a useful tool when performing research on a specific topic. 

After clicking you will see articles from the same authors and with the same keywords.

You can select the jurisdiction from either the search results page or the home page as well; simply click "select courts". You can also refine your search by state courts or federal courts. 

To quickly search a frequently used selection of courts, bookmark a search results page with the desired selection. 

 How do I sign up for email alerts?

Do a search for the topic of interest, e.g., "M Theory"; click the envelope icon in the sidebar of the search  results page; enter your email address, and click " Create alert ". Google will periodically email you newly published papers that match your search criteria. You can use any email address for this; it does not need to be a Google Account. 

If you want to get alerts from new articles published in a specific journal; type in the name of this journal in the search bar and create an alert like you would a keyword. 

How do I get notified of new papers published by my colleagues, advisors or professors?

First, do a search for your their name, and see if they have a Citations profile. If they do, click on it, and click the "Follow new articles" link in the right sidebar under the search box.

If they don't have a profile, do a search by author, e.g., [author:s-hawking], and click on the mighty envelope in the left sidebar of the search results page. If you find that several different people share the same name, you may need to add co-author names or topical keywords to limit results to the author you wish to follow.

How do I change my alerts?

If you created alerts using a Google account, you can manage them all on the "Alerts" page . 

From here you can create, edit or delete alerts. Select cancel under the actions column to unsubscribe from an alert. 

This will pop-open the advanced search menu

Here you can search specific words/phrases as well as for author, title and journal. You can also limit your search results by date.

• << Previous: Enable My Library
• Next: Google Scholar Library >>
• Last Updated: Feb 14, 2024 2:55 PM
• URL: https://libguides.uml.edu/googlescholar

• Harvard Library
• Research Guides
• Faculty of Arts & Sciences Libraries

A Scholar's Guide to Google

• Google Scholar
• Google Books

Using Google Scholar

Google Scholar is a special version of Google specially designed for searching scholarly literature. It covers peer-reviewed papers, theses, books, preprints, abstracts and technical reports from all broad areas of research.

A Harvard ID and PIN are required for Google Scholar in order to access the full text of books, journal articles, etc. provided by licensed resources to which Harvard subscribes. Indviduals outside of Harvard may access Google Scholar directly at http://scholar.google.com/ , but they will not have access to the full text of articles provided by Harvard Library E-Resources .

Browsing Search Results

The following screenshots illustrate some of the features that accompany individual records in Google Scholar's results lists.

Find It@Harvard – Locates an electronic version of the work (when available) through Harvard's subscription library resources. If no electronic full text is available, a link to the appropriate HOLLIS Catalog record is provided for alternative formats.

Group of – Finds other articles included in this group of scholarly works, possibly preliminary, which you may be able to access. Examples include preprints, abstracts, conference papers or other adaptations.

Cited By – Identifies other papers that have cited articles in the group.

Related Articles - The list of related articles is ranked primarily by how similar these articles are to the original result, but also takes into account the relevance of each paper. Finding sets of related papers and books is often a great way for novices to get acquainted with a topic.

Cached - The "Cached" link is the snapshot that Google took of the page when they crawled the web. The page may have changed since that time and the cached page may reference images which are no longer available.

Web Search – Searches for information on the Web about this work using the Google search engine.

BL Direct – Purchase the full text of the article through the British Library. Once transferred into BL Direct, users can also link to the full collection of The British Library document supply content. Prices for the service are expressed in British pounds. Abstracts for some documents are provided.

The Advanced Search feature in Google Scholar allows researchers to limit their query to particular authors, publications, dates, and subject areas.  

Page Last Reviewed: February 25, 2008

• << Previous: Google Books
• Last Updated: Jun 8, 2017 1:21 PM
• URL: https://guides.library.harvard.edu/googleguide

Harvard University Digital Accessibility Policy

Google Research: Themes from 2021 and Beyond

Over the last several decades, I've witnessed a lot of change in the fields of machine learning (ML) and computer science. Early approaches, which often fell short, eventually gave rise to modern approaches that have been very successful. Following that long-arc pattern of progress, I think we'll see a number of exciting advances over the next several years, advances that will ultimately benefit the lives of billions of people with greater impact than ever before. In this post, I’ll highlight five areas where ML is poised to have such impact. For each, I’ll discuss related research (mostly from 2021) and the directions and progress we’ll likely see in the next few years.

Trend 1: More Capable, General-Purpose ML Models

Researchers are training larger, more capable machine learning models than ever before. For example, just in the last couple of years models in the language domain have grown from billions of parameters trained on tens of billions of tokens of data (e.g., the 11B parameter T5 model), to hundreds of billions or trillions of parameters trained on trillions of tokens of data (e.g., dense models such as OpenAI’s 175B parameter GPT-3 model and DeepMind’s 280B parameter Gopher model, and sparse models such as Google’s 600B parameter GShard model and 1.2T parameter GLaM model). These increases in dataset and model size have led to significant increases in accuracy for a wide variety of language tasks, as shown by across-the-board improvements on standard natural language processing (NLP) benchmark tasks (as predicted by work on neural scaling laws for language models and machine translation models ).

Many of these advanced models are focused on the single but important modality of written language and have shown state-of-the-art results in language understanding benchmarks and open-ended conversational abilities, even across multiple tasks in a domain. They have also shown exciting capabilities to generalize to new language tasks with relatively little training data, in some cases, with few to no training examples for a new task . A couple of examples include improved long-form question answering , zero-label learning in NLP , and our LaMDA model, which demonstrates a sophisticated ability to carry on open-ended conversations that maintain significant context across multiple turns of dialog.

Transformer models are also having a major impact in image, video, and speech models, all of which also benefit significantly from scale, as predicted by work on scaling laws for visual transformer models . Transformers for image recognition and for video classification are achieving state-of-the-art results on many benchmarks, and we’ve also demonstrated that co-training models on both image data and video data can improve performance on video tasks compared with video data alone. We’ve developed sparse, axial attention mechanisms for image and video transformers that use computation more efficiently, found better ways of tokenizing images for visual transformer models , and improved our understanding of visual transformer methods by examining how they operate compared with convolutional neural networks . Combining transformer models with convolutional operations has shown significant benefits in visual as well as speech recognition tasks.

The outputs of generative models are also substantially improving. This is most apparent in generative models for images, which have made significant strides over the last few years. For example, recent models have demonstrated the ability to create realistic images given just a category (e.g., "irish setter" or "streetcar", if you desire), can "fill in" a low-resolution image to create a natural-looking high-resolution counterpart ("computer, enhance!"), and can even create natural-looking aerial nature scenes of arbitrary length . As another example, images can be converted to a sequence of discrete tokens that can then be synthesized at high fidelity with an autoregressive generative model.

Because these are powerful capabilities that come with great responsibility, we carefully vet potential applications of these sorts of models against our AI Principles .

Beyond advanced single-modality models, we are also starting to see large-scale multi-modal models. These are some of the most advanced models to date because they can accept multiple different input modalities (e.g., language, images, speech, video) and, in some cases, produce different output modalities, for example, generating images from descriptive sentences or paragraphs , or describing the visual content of images in human languages . This is an exciting direction because like the real world, some things are easier to learn in data that is multimodal (e.g., reading about something and seeing a demonstration is more useful than just reading about it). As such, pairing images and text can help with multi-lingual retrieval tasks , and better understanding of how to pair text and image inputs can yield improved results for image captioning tasks. Similarly, jointly training on visual and textual data can also help improve accuracy and robustness on visual classification tasks, while co-training on image, video, and audio tasks improves generalization performance for all modalities . There are also tantalizing hints that natural language can be used as an input for image manipulation , telling robots how to interact with the world and controlling other software systems, portending potential changes to how user interfaces are developed. Modalities handled by these models will include speech, sounds, images, video, and languages, and may even extend to structured data , knowledge graphs , and time series data .

Often these models are trained using self-supervised learning approaches, where the model learns from observations of “raw” data that has not been curated or labeled, e.g., language models used in GPT-3 and GLaM , the self-supervised speech model BigSSL , the visual contrastive learning model SimCLR , and the multimodal contrastive model VATT . Self-supervised learning allows a large speech recognition model to match the previous Voice Search automatic speech recognition (ASR) benchmark accuracy while using only 3% of the annotated training data. These trends are exciting because they can substantially reduce the effort required to enable ML for a particular task, and because they make it easier (though by no means trivial) to train models on more representative data that better reflects different subpopulations, regions, languages, or other important dimensions of representation.

All of these trends are pointing in the direction of training highly capable general-purpose models that can handle multiple modalities of data and solve thousands or millions of tasks. By building in sparsity, so that the only parts of a model that are activated for a given task are those that have been optimized for it, these multimodal models can be made highly efficient. Over the next few years, we are pursuing this vision in a next-generation architecture and umbrella effort called Pathways . We expect to see substantial progress in this area, as we combine together many ideas that to date have been pursued relatively independently.

Trend 2: Continued Efficiency Improvements for ML

Improvements in efficiency — arising from advances in computer hardware design as well as ML algorithms and meta-learning research — are driving greater capabilities in ML models. Many aspects of the ML pipeline, from the hardware on which a model is trained and executed to individual components of the ML architecture, can be optimized for efficiency while maintaining or improving on state-of-the-art performance overall. Each of these different threads can improve efficiency by a significant multiplicative factor, and taken together, can reduce computational costs, including CO 2 equivalent emissions ( CO2e ), by orders of magnitude compared to just a few years ago. This greater efficiency has enabled a number of critical advances that will continue to dramatically improve the efficiency of machine learning, enabling larger, higher quality ML models to be developed cost effectively and further democratizing access. I’m very excited about these directions of research!

Continued Improvements in ML Accelerator Performance

Each generation of ML accelerator improves on previous generations, enabling faster performance per chip, and often increasing the scale of the overall systems. Last year, we announced our TPUv4 systems , the fourth generation of Google’s Tensor Processing Unit, which demonstrated a 2.7x improvement over comparable TPUv3 results in the MLPerf benchmarks . Each TPUv4 chip has ~2x the peak performance per chip versus the TPUv3 chip, and the scale of each TPUv4 pod is 4096 chips (4x that of TPUv3 pods), yielding a performance of approximately 1.1 exaflops per pod (versus ~100 petaflops per TPUv3 pod). Having pods with larger numbers of chips that are connected together with high speed networks improves efficiency for larger models. ML capabilities on mobile devices are also increasing significantly. The Pixel 6 phone features a brand new Google Tensor processor that integrates a powerful ML accelerator to better support important on-device features . Left: TPUv4 board; Center: Part of a TPUv4 pod; Right: Google Tensor chip found in Pixel 6 phones. Our use of ML to accelerate the design of computer chips of all kinds (more on this below) is also paying dividends, particularly to produce better ML accelerators.

Continued Improvements in ML Compilation and Optimization of ML Workloads

Even when the hardware is unchanged, improvements in compilers and other optimizations in system software for machine learning accelerators can lead to significant improvements in efficiency. For example, “ A Flexible Approach to Autotuning Multi-pass Machine Learning Compilers ” shows how to use machine learning to perform auto-tuning of compilation settings to get across-the-board performance improvements of 5-15% (and sometimes as much as 2.4x improvement) for a suite of ML programs on the same underlying hardware. GSPMD describes an automatic parallelization system based on the XLA compiler that is capable of scaling most deep learning network architectures beyond the memory capacity of an accelerator and has been applied to many large models, such as GShard-M4 , LaMDA , BigSSL , ViT , MetNet-2 , and GLaM , leading to state-of-the-art results across several domains. End-to-end model speedups from using ML-based compiler autotuning on 150 ML models. Included are models that achieve improvements of 5% or more. Bar colors represent relative improvement from optimizing different model components.

Human-Creativity–Driven Discovery of More Efficient Model Architectures

Continued improvements in model architectures give substantial reductions in the amount of computation needed to achieve a given level of accuracy for many problems. For example, the Transformer architecture, which we developed in 2017 , was able to improve the state of the art on several NLP and translation benchmarks while simultaneously using 10x to 100x less computation to achieve these results than a variety of other prevalent methods, such as LSTMs and other recurrent architectures. Similarly, the Vision Transformer was able to show improved state-of-the-art results on a number of different image classification tasks despite using 4x to 10x less computation than convolutional neural networks.

Machine-Driven Discovery of More Efficient Model Architectures

Neural architecture search (NAS) can automatically discover new ML architectures that are more efficient for a given problem domain. A primary advantage of NAS is that it can greatly reduce the effort needed for algorithm development, because NAS requires only a one-time effort per search space and problem domain combination. In addition, while the initial effort to perform NAS can be computationally expensive, the resulting models can greatly reduce computation in downstream research and production settings, resulting in greatly reduced resource requirements overall. For example, the one-time search to discover the Evolved Transformer generated only 3.2 tons of CO2e (much less than the 284t CO 2 e reported elsewhere ; see Appendix C and D in this joint Google/UC Berkeley preprint ), but yielded a model for use by anyone in the NLP community that is 15-20% more efficient than the plain Transformer model. A more recent use of NAS discovered an even more efficient architecture called Primer (that has also been open-sourced ), which reduces training costs by 4x compared to a plain Transformer model. In this way, the discovery costs of NAS searches are often recouped from the use of the more-efficient model architectures that are discovered, even if they are applied to only a handful of downstream uses (and many NAS results are reused thousands of times). The Primer architecture discovered by NAS is 4x as efficient compared with a plain Transformer model. This image shows (in red) the two main modifications that give Primer most of its gains: depthwise convolution added to attention multi-head projections and squared ReLU activations (blue indicates portions of the original Transformer). NAS has also been used to discover more efficient models in the vision domain. The EfficientNetV2 model architecture is the result of a neural architecture search that jointly optimizes for model accuracy, model size, and training speed. On the ImageNet benchmark, EfficientNetV2 improves training speed by 5–11x while substantially reducing model size over previous state-of-the-art models. The CoAtNet model architecture was created with an architecture search that uses ideas from the Vision Transformer and convolutional networks to create a hybrid model architecture that trains 4x faster than the Vision Transformer and achieves a new ImageNet state of the art. EfficientNetV2 achieves much better training efficiency than prior models for ImageNet classification. The broad use of search to help improve ML model architectures and algorithms, including the use of reinforcement learning and evolutionary techniques, has inspired other researchers to apply this approach to different domains. To aid others in creating their own model searches, we have open-sourced Model Search , a platform that enables others to explore model search for their domains of interest. In addition to model architectures, automated search can also be used to find new, more efficient reinforcement learning algorithms , building on the earlier AutoML-Zero work that demonstrated this approach for automating supervised learning algorithm discovery.

Use of Sparsity

Sparsity, where a model has a very large capacity, but only some parts of the model are activated for a given task, example or token, is another important algorithmic advance that can greatly improve efficiency. In 2017, we introduced the sparsely-gated mixture-of-experts layer , which demonstrated better results on a variety of translation benchmarks while using 10x less computation than previous state-of-the-art dense LSTM models. More recently, Switch Transformers , which pair a mixture-of-experts –style architecture with the Transformer model architecture, demonstrated a 7x speedup in training time and efficiency over the dense T5-Base Transformer model. The GLaM model showed that transformers and mixture-of-expert–style layers can be combined to produce a model that exceeds the accuracy of the GPT-3 model on average across 29 benchmarks using 3x less energy for training and 2x less computation for inference. The notion of sparsity can also be applied to reduce the cost of the attention mechanism in the core Transformer architecture . The BigBird sparse attention model consists of global tokens that attend to all parts of an input sequence, local tokens, and a set of random tokens. Theoretically, this can be interpreted as adding a few global tokens on a Watts-Strogatz graph . The use of sparsity in models is clearly an approach with very high potential payoff in terms of computational efficiency, and we are only scratching the surface in terms of research ideas to be tried in this direction. Each of these approaches for improved efficiency can be combined together so that equivalent-accuracy language models trained today in efficient data centers are ~100 times more energy efficient and produce ~650 times less CO 2 e emissions, compared to a baseline Transformer model trained using P100 GPUs in an average U.S. datacenter using an average U.S. energy mix. And this doesn’t even account for Google’s carbon-neutral, 100% renewable energy offsets. We’ll have a more detailed blog post analyzing the carbon emissions trends of NLP models soon.

Trend 3: ML Is Becoming More Personally and Communally Beneficial

A host of new experiences are made possible as innovation in ML and silicon hardware (like the Google Tensor processor on the Pixel 6) enable mobile devices to be more capable of continuously and efficiently sensing their surrounding context and environment. These advances have improved accessibility and ease of use, while also boosting computational power, which is critical for popular features like mobile photography, live translation and more. Remarkably, recent technological advances also provide users with a more customized experience while strengthening privacy safeguards.

More people than ever rely on their phone cameras to record their daily lives and for artistic expression. The clever application of ML to computational photography has continued to advance the capabilities of phone cameras, making them easier to use, improving performance, and resulting in higher-quality images. Advances, such as improved HDR+ , the ability to take pictures in very low light , better handling of portraits , and efforts to make cameras more inclusive so they work for all skin tones , yield better photos that are more true to the photographer’s vision and to their subjects. Such photos can be further improved using the powerful ML-based tools now available in Google Photos, like cinematic photos , noise and blur reduction , and the Magic Eraser .

In addition to using their phones for creative expression, many people rely on them to help communicate with others across languages and modalities in real-time using Live Translate in messaging apps and Live Caption for phone calls . Speech recognition accuracy has continued to make substantial improvements thanks to techniques like self-supervised learning and noisy student training , with marked improvements for accented speech, noisy conditions or environments with overlapping speech , and across many languages. Building on advances in text-to-speech synthesis, people can listen to web pages and articles using our Read Aloud technology on a growing number of platforms , making information more available across barriers of modality and languages. Live speech translations in the Google Translate app have become significantly better by stabilizing the translations that are generated on-the-fly, and high quality, robust and responsible direct speech-to-speech translation provides a much better user experience in communicating with people speaking a different language. New work on combining ML with traditional codec approaches in the Lyra speech codec and the more general SoundStream audio codec enables higher fidelity speech, music, and other sounds to be communicated reliably at much lower bitrate.

Everyday interactions are becoming much more natural with features like automatic call screening and ML agents that will wait on hold for you , thanks to advances in Duplex . Even short tasks that users may perform frequently have been improved with tools such as Smart Text Selection , which automatically selects entities like phone numbers or addresses for easy copy and pasting, and grammar correction as you type on Pixel 6 phones. In addition, Screen Attention prevents the phone screen from dimming when you are looking at it and improvements in gaze recognition are opening up new use cases for accessibility and for improved wellness and health . ML is also enabling new methods for ensuring the safety of people and communities. For example, Suspicious Message Alerts warn against possible phishing attacks and Safer Routing detects hard-braking events to suggest alternate routes.

Given the potentially sensitive nature of the data that underlies these new capabilities, it is essential that they are designed to be private by default. Many of them run inside of Android's Private Compute Core — an open source, secure environment isolated from the rest of the operating system. Android ensures that data processed in the Private Compute Core is not shared to any apps without the user taking an action. Android also prevents any feature inside the Private Compute Core from having direct access to the network. Instead, features communicate over a small set of open-source APIs to Private Compute Services , which strips out identifying information and makes use of privacy technologies, including federated learning , federated analytics , and private information retrieval, enabling learning while simultaneously ensuring privacy.

These technologies are critical to evolving next-generation computation and interaction paradigms, whereby personal or communal devices can both learn from and contribute to training a collective model of the world without compromising privacy. A federated unsupervised approach to privately learn the kinds of aforementioned general-purpose models with fine-tuning for a given task or context could unlock increasingly intelligent systems that are far more intuitive to interact with — more like a social entity than a machine. Broad and equitable access to these intelligent interfaces will only be possible with deep changes to our technology stacks, from the edge to the datacenter, so that they properly support neural computing.

Trend 4: Growing Impact of ML in Science, Health and Sustainability

In recent years, we have seen an increasing impact of ML in the basic sciences, from physics to biology, with a number of exciting practical applications in related realms, such as renewable energy and medicine. Computer vision models have been deployed to address problems at both personal and global scales. They can assist physicians in their regular work, expand our understanding of neural physiology, and also provide better weather forecasts and streamline disaster relief efforts. Other types of ML models are proving critical in addressing climate change by discovering ways to reduce emissions and improving the output of alternative energy sources. Such models can even be leveraged as creative tools for artists! As ML becomes more robust, well-developed, and widely accessible, its potential for high-impact applications in a broad array of real-world domains continues to expand, helping to solve some of our most challenging problems.

Large-Scale Application of Computer Vision for New Insights

The advances in computer vision over the past decade have enabled computers to be used for a wide variety of tasks across different scientific domains. In neuroscience, automated reconstruction techniques can recover the neural connective structure of brain tissues from high resolution electron microscopy images of thin slices of brain tissue. In previous years, we have collaborated to create such resources for fruit fly , mouse, and songbird brains, but last year, we collaborated with the Lichtman Lab at Harvard University to analyze the largest sample of brain tissue imaged and reconstructed in this level of detail, in any species, and produced the first large-scale study of synaptic connectivity in the human cortex that spans multiple cell types across all layers of the cortex. The goal of this work is to produce a novel resource to assist neuroscientists in studying the stunning complexity of the human brain. The image below, for example, shows six neurons out of about 86 billion neurons in an adult human brain . A single human chandelier neuron from our human cortex reconstruction, along with some of the pyramidal neurons that make a connection with that cell. Here’s an interactive version and a gallery of other interactive examples . Computer vision technology also provides powerful tools to address challenges at much larger, even global, scales. A deep-learning–based approach to weather forecasting that uses satellite and radar imagery as inputs, combined with other atmospheric data, produces weather and precipitation forecasts that are more accurate than traditional physics-based models at forecasting times up to 12 hours. They can also produce updated forecasts much more quickly than traditional methods, which can be critical in times of extreme weather. Comparison of 0.2 mm/hr precipitation on March 30, 2020 over Denver, Colorado. Left: Ground truth, source MRMS . Center: Probability map as predicted by MetNet-2 . Right: Probability map as predicted by the physics-based HREF model. MetNet-2 is able to predict the onset of the storm earlier in the forecast than HREF as well as the storm’s starting location, whereas HREF misses the initiation location, but captures its growth phase well. Having an accurate record of building footprints is essential for a range of applications, from population estimation and urban planning to humanitarian response and environmental science. In many parts of the world, including much of Africa, this information wasn’t previously available, but new work shows that using computer vision techniques applied to satellite imagery can help identify building boundaries at continental scales . The results of this approach have been released in the Open Buildings dataset , a new open-access data resource that contains the locations and footprints of 516 million buildings with coverage across most of the African continent. We’ve also been able to use this unique dataset in our collaboration with the World Food Programme to provide fast damage assessment after natural disasters through application of ML. Example of segmenting buildings in satellite imagery. Left: Source image; Center: Semantic segmentation, with each pixel assigned a confidence score that it is a building vs. non-building; Right: Instance segmentation, obtained by thresholding and grouping together connected components. A common theme across each of these cases is that ML models are able to perform specialized tasks efficiently and accurately based on analysis of available visual data, supporting high impact downstream tasks.

Automated Design Space Exploration

Another approach that has yielded excellent results across many fields is to allow an ML algorithm to explore and evaluate a problem’s design space for possible solutions in an automated way. In one application, a Transformer-based variational autoencoder learns to create aesthetically-pleasing and useful document layouts , and the same approach can be extended to explore possible furniture layouts. Another ML-driven approach automates the exploration of the huge design space of tweaks for computer game rules to improve playability and other attributes of a game, enabling human game designers to create enjoyable games more quickly. A visualization of the Variational Transformer Network (VTN) model, which is able to extract meaningful relationships between the layout elements (paragraphs, tables, images, etc.) in order to generate realistic synthetic documents (e.g., with better alignment and margins). Other ML algorithms have been used to evaluate the design space of computer architectural decisions for ML accelerator chips themselves. We’ve also shown that ML can be used to quickly create chip placements for ASIC designs that are better than layouts generated by human experts and can be generated in a matter of hours instead of weeks. This reduces the fixed engineering costs of chips and lowers the barrier to quickly creating specialized hardware for different applications. We’ve successfully used this automated placement approach in the design of our upcoming TPU-v5 chip. Such exploratory ML approaches have also been applied to materials discovery. In a collaboration between Google Research and Caltech, several ML models, combined with a modified inkjet printer and a custom-built microscope, were able to rapidly search over hundreds of thousands of possible materials to hone in on 51 previously uncharacterized three-metal oxide materials with promising properties for applications in areas like battery technology and electrolysis of water. These automated design space exploration approaches can help accelerate many scientific fields, especially when the entire experimental loop of generating the experiment and evaluating the result can all be done in an automated or mostly-automated manner. I expect to see this approach applied to good effect in many more areas in the coming years.

Application to Health

In addition to advancing basic science, ML can also drive advances in medicine and human health more broadly. The idea of leveraging advances in computer science in health is nothing new — in fact some of my own early experiences were in developing software to help analyze epidemiological data . But ML opens new doors, raises new opportunities, and yes, poses new challenges. Take for example the field of genomics. Computing has been important to genomics since its inception, but ML adds new capabilities and disrupts old paradigms. When Google researchers began working in this area, the idea of using deep learning to help infer genetic variants from sequencer output was considered far-fetched by many experts. Today, this ML approach is considered state-of-the-art . But the future holds an even more important role for ML — genomics companies are developing new sequencing instruments that are more accurate and faster, but also present new inference challenges. Our release of open-source software DeepConsensus and, in collaboration with UCSC, PEPPER-DeepVariant , supports these new instruments with cutting-edge informatics. We hope that more rapid sequencing can lead to near term applicability with impact for real patients. A schematic of the Transformer architecture for DeepConsensus , which corrects sequencing errors to improve yield and correctness. There are other opportunities to use ML to accelerate our use of genomic information for personalized health outside of processing the sequencer data. Large biobanks of extensively phenotyped and sequenced individuals can revolutionize how we understand and manage genetic predisposition to disease. Our ML-based phenotyping method improves the scalability of converting large imaging and text datasets into phenotypes usable for genetic association studies, and our DeepNull method better leverages large phenotypic data for genetic discovery. We are happy to release both as open-source methods for the scientific community. The process for generating large-scale quantification of anatomical and disease traits for combination with genomic data in Biobanks. Just as ML helps us see hidden characteristics of genomics data, it can help us discover new information and glean new insights from other health data types as well. Diagnosis of disease is often about identifying a pattern, quantifying a correlation, or recognizing a new instance of a larger class — all tasks at which ML excels. Google researchers have used ML to tackle a wide range of such problems , but perhaps none of these has progressed farther than the applications of ML to medical imaging. In fact, Google’s 2016 paper describing the application of deep learning to the screening for diabetic retinopathy , was selected by the editors of the Journal of the American Medical Association (JAMA) as one of the top 10 most influential papers of the decade — not just the most influential papers on ML and health, the most influential JAMA papers of the decade overall. But the strength of our research doesn’t end at contributions to the literature, but extends to our ability to build systems operating in the real world. Through our global network of deployment partners, this same program has helped screen tens of thousands of patients in India, Thailand, Germany and France who might otherwise have been untested for this vision-threatening disease. We expect to see this same pattern of assistive ML systems deployed to improve breast cancer screening , detect lung cancer , accelerate radiotherapy treatments for cancer , flag abnormal X-rays , and stage prostate cancer biopsies . Each domain presents new opportunities to be helpful. ML-assisted colonoscopy procedures are a particularly interesting example of going beyond the basics. Colonoscopies are not just used to diagnose colon cancer — the removal of polyps during the procedure are the front line of halting disease progression and preventing serious illness. In this domain we’ve demonstrated that ML can help ensure doctors don’t miss polyps , can help detect elusive polyps , and can add new dimensions of quality assurance, like coverage mapping through the application of simultaneous localization and mapping techniques. In collaboration with Shaare Zedek Medical Center in Jerusalem, we’ve shown these systems can work in real time, detecting an average of one polyp per procedure that would have otherwise been missed, with fewer than four false alarms per procedure. Sample chest X-rays (CXR) of true and false positives, and true and false negatives for ( A ) general abnormalities, ( B ) tuberculosis, and ( C ) COVID-19. On each CXR, red outlines indicate areas on which the model focused to identify abnormalities (i.e., the class activation map), and yellow outlines refer to regions of interest identified by a radiologist. Another ambitious healthcare initiative, Care Studio, uses state-of-the-art ML and advanced NLP techniques to analyze structured data and medical notes , presenting clinicians with the most relevant information at the right time — ultimately helping them deliver more proactive and accurate care. As important as ML may be to expanding access and improving accuracy in the clinical setting, we see a new equally important trend emerging: ML applied to help people in their daily health and well-being. Our everyday devices have powerful sensors that can help democratize health metrics and information so people can make more informed decisions about their health. We’ve already seen launches that enable a smartphone camera to assess heart rate and respiratory rate to help users without additional hardware, and Nest Hub devices that support contactless sleep sensing and allow users to better understand their nighttime wellness. We’ve seen that we can, on the one hand, significantly improve speech recognition quality for disordered speech in our own ASR systems, and on the other, use ML to help recreate the voice of those with speech impairments , empowering them to communicate in their own voice. ML enabled smartphones that help people better research emerging skin conditions or help those with limited vision go for a jog , seem to be just around the corner. These opportunities offer a future too bright to ignore. The custom ML model for contactless sleep sensing efficiently processes a continuous stream of 3D radar tensors (summarizing activity over a range of distances, frequencies, and time) to automatically compute probabilities for the likelihood of user presence and wakefulness (awake or asleep).

ML Applications for the Climate Crisis

Another realm of paramount importance is climate change, which is an incredibly urgent threat for humanity. We need to all work together to bend the curve of harmful emissions to ensure a safe and prosperous future. Better information about the climate impact of different choices can help us tackle this challenge in a number of different ways . To this end, we recently rolled out eco-friendly routing in Google Maps , which we estimate will save about 1 million tons of CO 2 emissions per year (the equivalent of removing more than 200,000 cars from the road). A recent case study shows that using Google Maps directions in Salt Lake City results in both faster and more emissions-friendly routing, which saves 1.7% of CO 2 emissions and 6.5% travel time. In addition, making our Maps products smarter about electric vehicles can help alleviate range anxiety, encouraging people to switch to emissions-free vehicles. We are also working with multiple municipalities around the world to use aggregated historical traffic data to help suggest improved traffic light timing settings, with an early pilot study in Israel and Brazil showing a 10-20% reduction in fuel consumption and delay time at the examined intersections. With eco-friendly routing, Google Maps will show you the fastest route and the one that’s most fuel-efficient — so you can choose whichever one works best for you. On a longer time scale, fusion holds promise as a game-changing renewable energy source. In a long-standing collaboration with TAE Technologies, we have used ML to help maintain stable plasmas in their fusion reactor by suggesting settings of the more than 1000 relevant control parameters. With our collaboration, TAE achieved their major goals for their Norman reactor, which brings us a step closer to the goal of breakeven fusion . The machine maintains a stable plasma at 30 million Kelvin (don’t touch!) for 30 milliseconds, which is the extent of available power to its systems. They have completed a design for an even more powerful machine, which they hope will demonstrate the conditions necessary for breakeven fusion before the end of the decade. We’re also expanding our efforts to address wildfires and floods, which are becoming more common (like millions of Californians, I’m having to adapt to having a regular “fire season”). Last year, we launched a wildfire boundary map powered by satellite data to help people in the U.S. easily understand the approximate size and location of a fire — right from their device. Building on this, we’re now bringing all of Google’s wildfire information together and launching it globally with a new layer on Google Maps. We have been applying graph optimization algorithms to help optimize fire evacuation routes to help keep people safe in the presence of rapidly advancing fires. In 2021, our Flood Forecasting Initiative expanded its operational warning systems to cover 360 million people, and sent more than 115 million notifications directly to the mobile devices of people at risk from flooding, more than triple our outreach in the previous year. We also deployed our LSTM-based forecast models and the new Manifold inundation model in real-world systems for the first time, and shared a detailed description of all components of our systems . The wildfire layer in Google Maps provides people with critical, up-to-date information in an emergency. We’re also working hard on our own set of sustainability initiatives. Google was the first major company to become carbon neutral in 2007. We were also the first major company to match our energy use with 100 percent renewable energy in 2017. We operate the cleanest global cloud in the industry, and we’re the world’s largest corporate purchaser of renewable energy. Further, in 2020 we became the first major company to make a commitment to operate on 24/7 carbon-free energy in all our data centers and campuses worldwide. This is far more challenging than the traditional approach of matching energy usage with renewable energy, but we’re working to get this done by 2030. Carbon emission from ML model training is a concern for the ML community, and we have shown that making good choices about model architecture, datacenter, and ML accelerator type can reduce the carbon footprint of training by ~100-1000x.

Trend 5: Deeper and Broader Understanding of ML

As ML is used more broadly across technology products and society more generally, it is imperative that we continue to develop new techniques to ensure that it is applied fairly and equitably, and that it benefits all people and not just select subsets. This is a major focus for our Responsible AI and Human-Centered Technology research group and an area in which we conduct research on a variety of responsibility-related topics .

One area of focus is recommendation systems that are based on user activity in online products. Because these recommendation systems are often composed of multiple distinct components, understanding their fairness properties often requires insight into individual components as well as how the individual components behave when combined together. Recent work has helped to better understand these relationships , revealing ways to improve the fairness of both individual components and the overall recommendation system. In addition, when learning from implicit user activity, it is also important for recommendation systems to learn in an unbiased manner , since the straightforward approach of learning from items that were shown to previous users exhibits well-known forms of bias. Without correcting for such biases, for example, items that were shown in more prominent positions to users tend to get recommended to future users more often.

As in recommendation systems, surrounding context is important in machine translation. Because most machine translation systems translate individual sentences in isolation, without additional surrounding context, they can often reinforce biases related to gender, age or other areas. In an effort to address some of these issues, we have a long-standing line of research on reducing gender bias in our translation systems , and to help the entire translation community, last year we released a dataset to study gender bias in translation based on translations of Wikipedia biographies.

Another common problem in deploying machine learning models is distributional shift: if the statistical distribution of data on which the model was trained is not the same as that of the data the model is given as input, the model’s behavior can sometimes be unpredictable. In recent work, we employ the Deep Bootstrap framework to compare the real world, where there is finite training data, to an "ideal world", where there is infinite data. Better understanding of how a model behaves in these two regimes (real vs. ideal) can help us develop models that generalize better to new settings and exhibit less bias towards fixed training datasets.

Although work on ML algorithms and model development gets significant attention, data collection and dataset curation often gets less. But this is an important area, because the data on which an ML model is trained can be a potential source of bias and fairness issues in downstream applications. Analyzing such data cascades in ML can help identify the many places in the lifecycle of an ML project that can have substantial influence on the outcomes. This research on data cascades has led to evidence-backed guidelines for data collection and evaluation in the revised PAIR Guidebook , aimed at ML developers and designers.

The general goal of better understanding data is an important part of ML research. One thing that can help is finding and investigating anomalous data . We have developed methods to better understand the influence that particular training examples can have on an ML model, since mislabeled data or other similar issues can have outsized impact on the overall model behavior. We have also built the Know Your Data tool to help ML researchers and practitioners better understand properties of their datasets, and last year we created a case study of how to use the Know Your Data tool to explore issues like gender bias and age bias in a dataset.

Understanding dynamics of benchmark dataset usage is also important, given the central role they play in the organization of ML as a field. Although studies of individual datasets have become increasingly common, the dynamics of dataset usage across the field have remained underexplored. In recent work, we published the first large scale empirical analysis of dynamics of dataset creation, adoption, and reuse . This work offers insights into pathways to enable more rigorous evaluations, as well as more equitable and socially informed research.

Creating public datasets that are more inclusive and less biased is an important way to help improve the field of ML for everyone. In 2016, we released the Open Images dataset , a collection of ~9 million images annotated with image labels spanning thousands of object categories and bounding box annotations for 600 classes. Last year, we introduced the More Inclusive Annotations for People (MIAP) dataset in the Open Images Extended collection. The collection contains more complete bounding box annotations for the person class hierarchy, and each annotation is labeled with fairness-related attributes, including perceived gender presentation and perceived age range. With the increasing focus on reducing unfair bias as part of responsible AI research , we hope these annotations will encourage researchers already leveraging the Open Images dataset to incorporate fairness analysis in their research.

Because we also know that our teams are not the only ones creating datasets that can improve machine learning, we have built Dataset Search to help users discover new and useful datasets, wherever they might be on the Web.

Tackling various forms of abusive behavior online, such as toxic language, hate speech, and misinformation, is a core priority for Google. Being able to detect such forms of abuse reliably, efficiently, and at scale is of critical importance both to ensure that our platforms are safe and also to avoid the risk of reproducing such negative traits through language technologies that learn from online discourse in an unsupervised fashion. Google has pioneered work in this space through the Perspective API tool, but the nuances involved in detecting toxicity at scale remains a complex problem. In recent work, in collaboration with various academic partners, we introduced a comprehensive taxonomy to reason about the changing landscape of online hate and harassment . We also investigated how to detect covert forms of toxicity , such as microaggressions, that are often ignored in online abuse interventions, studied how conventional approaches to deal with disagreements in data annotations of such subjective concepts might marginalize minority perspectives , and proposed a new disaggregated modeling approach that uses a multi-task framework to tackle this issue. Furthermore, through qualitative research and network-level content analysis, Google’s Jigsaw team, in collaboration with researchers at George Washington University, studied how hate clusters spread disinformation across social media platforms .

Another potential concern is that ML language understanding and generation models can sometimes also produce results that are not properly supported by evidence. To confront this problem in question answering, summarization, and dialog, we developed a new framework for measuring whether results can be attributed to specific sources . We released annotation guidelines and demonstrated that they can be reliably used in evaluating candidate models.

Interactive analysis and debugging of models remains key to responsible use of ML. We have updated our Language Interpretability Tool with new capabilities and techniques to advance this line of work, including support for image and tabular data, a variety of features carried over from our previous work on the What-If Tool , and built-in support for fairness analysis through the technique of Testing with Concept Activation Vectors . Interpretability and explainability of ML systems more generally is also a key part of our Responsible AI vision ; in collaboration with DeepMind, we made headway in understanding the acquisition of human chess concepts in the self-trained AlphaZero chess system.

We are also working hard to broaden the perspective of Responsible AI beyond western contexts. Our recent research examines how various assumptions of conventional algorithmic fairness frameworks based on Western institutions and infrastructures may fail in non-Western contexts and offers a pathway for recontextualizing fairness research in India along several directions. We are actively conducting survey research across several continents to better understand perceptions of and preferences regarding AI . Western framing of algorithmic fairness research tends to focus on only a handful of attributes, thus biases concerning non-Western contexts are largely ignored and empirically under-studied. To address this gap, in collaboration with the University of Michigan, we developed a weakly supervised method to robustly detect lexical biases in broader geo-cultural contexts in NLP models that reflect human judgments of offensive and inoffensive language in those geographic contexts.

Furthermore, we have explored applications of ML to contexts valued in the Global South , including developing a proposal for farmer-centered ML research . Through this work, we hope to encourage the field to be thoughtful about how to bring ML-enabled solutions to smallholder farmers in ways that will improve their lives and their communities.

Involving community stakeholders at all stages of the ML pipeline is key to our efforts to develop and deploy ML responsibly and keep us focused on tackling the problems that matter most. In this vein, we held a Health Equity Research Summit among external faculty, non-profit organization leads, government and NGO representatives, and other subject matter experts to discuss how to bring more equity into the entire ML ecosystem, from the way we approach problem-solving to how we assess the impact of our efforts.

Community-based research methods have also informed our approach to designing for digital wellbeing and addressing racial equity issues in ML systems , including improving our understanding of the experience of Black Americans using ASR systems . We are also listening to the public more broadly to learn how sociotechnical ML systems could help during major life events, such as by supporting family caregiving.

As ML models become more capable and have impact in many domains, the protection of the private information used in ML continues to be an important focus for research. Along these lines, some of our recent work addresses privacy in large models, both highlighting that training data can sometimes be extracted from large models and pointing to how privacy can be achieved in large models, e.g., as in differentially private BERT . In addition to the work on federated learning and analytics, mentioned above, we have also been enhancing our toolbox with other principled and practical ML techniques for ensuring differential privacy, for example private clustering , private personalization , private matrix completion , private weighted sampling , private quantiles , private robust learning of halfspaces, and in general, sample-efficient private PAC learning . Moreover, we have been expanding the set of privacy notions that can be tailored to different applications and threat models, including label privacy and user versus item level privacy .

Recognizing the value of open datasets to the general advancement of ML and related fields of research, we continue to grow our collection of open source datasets and resources and expand our global index of open datasets in Google Dataset Search . This year, we have released a number of datasets and tools across a range of research areas:

Research Community Interaction

To realize our goal for a more robust and comprehensive understanding of ML and related technologies, we actively engage with the broader research community. In 2021, we published over 750 papers , nearly 600 of which were presented at leading research conferences. Google Research sponsored over 150 conferences, and Google researchers contributed directly by serving on program committees and organizing workshops, tutorials and numerous other activities aimed at collectively advancing the field. To learn more about our contributions to some of the larger research conferences this year, please see our recent conference blog posts . In addition, we hosted 19 virtual workshops (like the 2021 Quantum Summer Symposium ), which allowed us to further engage with the academic community by generating new ideas and directions for the research field and advancing research initiatives.

In 2021, Google Research also directly supported external research with $59M in funding, including $23M through Research programs to faculty and students, and $20M in university partnerships and outreach. This past year, we introduced new funding and collaboration programs that support academics all over the world who are doing high impact research. We funded 86 early career faculty through our Research Scholar Program to support general advancements in science, and funded 34 faculty through our Award for Inclusion Research Program who are doing research in areas like accessibility, algorithmic fairness, higher education and collaboration, and participatory ML. In addition to the research we are funding, we welcomed 85 faculty and post-docs, globally, through our Visiting Researcher program , to come to Google and partner with us on exciting ideas and shared research challenges. We also selected a group of 74 incredibly talented PhD student researchers to receive Google PhD Fellowships and mentorship as they conduct their research.

As part of our ongoing racial equity commitments , making computer science (CS) research more inclusive continues to be a top priority for us. In 2021, we continued expanding our efforts to increase the diversity of Ph.D. graduates in computing. For example, the CS Research Mentorship Program (CSRMP), an initiative by Google Research to support students from historically marginalized groups (HMGs) in computing research pathways, graduated 590 mentees, 83% of whom self-identified as part of an HMG, who were supported by 194 Google mentors — our largest group to date! In October, we welcomed 35 institutions globally leading the way to engage 3,400+ students in computing research as part of the 2021 exploreCSR cohort. Since 2018, this program has provided faculty with funding, community, evaluation and connections to Google researchers in order to introduce students from HMGs to the world of CS research. We are excited to expand this program to more international locations in 2022.

We also continued our efforts to fund and partner with organizations to develop and support new pathways and approaches to broadening participation in computing research at scale. From working with alliances like the Computing Alliance of Hispanic-Serving Institutions (CAHSI) and CMD-IT Diversifying LEAdership in the Professoriate (LEAP) Alliance to partnering with university initiatives like UMBC’s Meyerhoff Scholars , Cornell University’s CSMore , Northeastern University’s Center for Inclusive Computing , and MIT’s MEnTorEd Opportunities in Research (METEOR), we are taking a community-based approach to materially increase the representation of marginalized groups in computing research.

In writing these retrospectives, I try to focus on new research work that has happened (mostly) in the past year while also looking ahead. In past years’ retrospectives, I’ve tried to be more comprehensive, but this time I thought it could be more interesting to focus on just a few themes. We’ve also done great  work in many other research areas that don’t fit neatly into these themes. If you’re interested, I encourage you to check out our research publications by area below or by year (and if you’re interested in quantum computing, our Quantum team recently wrote a retrospective of their work in 2021 ):

Research is often a multi-year journey to real-world impact. Early stage research work that happened a few years ago is now having a dramatic impact on Google’s products and across the world. Investments in ML hardware accelerators like TPUs and in software frameworks like TensorFlow and JAX have borne fruit. ML models are increasingly prevalent in many different products and features at Google because their power and ease of expression streamline experimentation and productionization of ML models in performance-critical environments. Research into model architectures to create Seq2Seq , Inception , EfficientNet , and Transformer or algorithmic research like batch normalization and distillation is driving progress in the fields of language understanding, vision, speech, and others. Basic capabilities like better language and visual understanding and speech recognition can be transformational, and as a result, these sorts of models are widely deployed for a wide variety of problems in many of our products including Search, Assistant, Ads, Cloud, Gmail, Maps, YouTube, Workspace, Android, Pixel, Nest, and Translate.

These are truly exciting times in machine learning and computer science. Continued improvement in computers’ ability to understand and interact with the world around them through language, vision, and sound opens up entire new frontiers of how computers can help people accomplish things in the world. The many examples of progress along the five themes outlined in this post are waypoints in a long-term journey!

Acknowledgements

Thanks to Alison Carroll, Alison Lentz, Andrew Carroll, Andrew Tomkins, Avinatan Hassidim, Azalia Mirhoseini, Barak Turovsky, Been Kim, Blaise Aguera y Arcas, Brennan Saeta, Brian Rakowski, Charina Chou, Christian Howard, Claire Cui, Corinna Cortes, Courtney Heldreth, David Patterson, Dipanjan Das, Ed Chi, Eli Collins, Emily Denton, Fernando Pereira, Genevieve Park, Greg Corrado, Ian Tenney, Iz Conroy, James Wexler, Jason Freidenfelds, John Platt, Katherine Chou, Kathy Meier-Hellstern, Kyle Vandenberg, Lauren Wilcox, Lizzie Dorfman, Marian Croak, Martin Abadi, Matthew Flegal, Meredith Morris, Natasha Noy, Negar Saei, Neha Arora, Paul Muret, Paul Natsev, Quoc Le, Ravi Kumar, Rina Panigrahy, Sanjiv Kumar, Sella Nevo, Slav Petrov, Sreenivas Gollapudi, Tom Duerig, Tom Small, Vidhya Navalpakkam, Vincent Vanhoucke, Vinodkumar Prabhakaran, Viren Jain, Yonghui Wu, Yossi Matias, and Zoubin Ghahramani for helpful feedback and contributions to this post, and to the entire Research and Health communities at Google for everyone’s contributions towards this work.

Internet Sources for Research Help Guide

• Online Library Resources
• Scholarly Resources
• Government Resources
• Primary Sources
• Critiquing Websites

Introduction

Google books, google scholar, advanced search, quick search tips.

Google can be a good resource for research, if it is used effectively.

There are specific techniques that you can use to be an effective Google searcher. Your job is to decide which Google database you should use in order to find the types of materials you need, as well as to create searches, or queries, that provide pertinent results.

On the left are the three main Google collections that are useful for doing scholarly research: "regular" Google, Google Books and Google Scholar. On the right are tips for search techniques that will help target your search in such a way that your search results should pertain to the subject you are researching, as well as to the type of material you need.

This is the "regular" Google that we all use.

This is the search engine that allows you to look through Google's huge collection of digitized full-text books.

Google Scholar collects and gives you access to a huge number of scholarly works, including full-text articles and books.

The Google Advanced Search page gives you the capability to create effective and efficient searches without having to use the Google short cuts.

With this search page, you can limit your search by language, file type (.jpg, .pdf, etc) and date range, as well as searching for similar pages, or websites from a particular geographical region.

Google has a help page for Advanced Search

Use Boolean search terms Boolean terms refer to: AND, OR, NOT

These words tell a database how to do the search. "And" combines search terms; "OR" searches for either one search term or another; "NOT" ignores a particular word. (To see a visual representation of Boolean words, go here .)

AND : Google uses an implied "and" between search terms. For example, when you search for maryland constitution what Google does is look for maryland and constitution .

OR : Google will only recognize OR when it is in capital letters. maryland OR virginia

NOT : Google uses the minus sign to exclude terms. maryland -virginia

Restrict the Domain. You can direct Google to look for particularly types of websites, such as government, military, non-profit or education. This is done by indicating what type of site you want, using the "site:" command such as:

shays' rebellion site:.edu

Use Quotes To find words in a web page or document in the exact same order, put quotes around them:

"Song of Solomon"

Exclude non-necessary words Use only those keywords that describe your topic:

How did Frederick Douglass affect the Civil War? should be Frederick Douglass Civil War or "Frederick Douglass" "Civil War"

Search synonyms You can search for the synonyms of words by putting a tilde in front of the search term:

~love would search for "marriage," "romantic," "romance," as well as "love."

Search singular and plural Google does not automatically search for the plural form of words. To makes sure it does you have to use the Boolean OR:

sculpture OR sculptures

Searching for common words Google ignores common words such as "how," "this," "where," "a." To make sure that Google does a search for a word like this use the + symbol before the word:

+who +are +you

Use the "fill in the blank" feature Google can still look for something even if you can't remember the full name, or don't know a specific date, etc:

roe v * would search for court cases that began with "roe."

More Support

Google also provides information on how to do searches:

Basic Search Help

Advanced Tips

• << Previous: Critiquing Websites

[email protected] 202-274-6120

Facebook Twitter Instagram YouTube

• {{link.text}}

Publications

Google publishes hundreds of research papers each year. Publishing is important to us; it enables us to collaborate and share ideas with, as well as learn from, the broader scientific community. Submissions are often made stronger by the fact that ideas have been tested through real product implementation by the time of publication.

We believe the formal structures of publishing today are changing - in computer science especially, there are multiple ways of disseminating information.  We encourage publication both in conventional scientific venues, and through other venues such as industry forums, standards bodies, and open source software and product feature releases.

Open Source

We understand the value of a collaborative ecosystem and love open source software .

Product and Feature Launches

With every launch, we're publishing progress and pushing functionality.

Industry Standards

Our researchers are often helping to define not just today's products but also tomorrow's.

"Resources" doesn't just mean tangible assets but also intellectual. Incredible datasets and a great team of colleagues foster a rich and collaborative research environment.

Couple big challenges with big resources and Google offers unprecedented research opportunities.

22 Research Areas

• Algorithms and Theory 608 Publications
• Data Management 116 Publications
• Data Mining and Modeling 214 Publications
• Distributed Systems and Parallel Computing 208 Publications
• Economics and Electronic Commerce 209 Publications
• Education Innovation 30 Publications
• General Science 158 Publications
• Hardware and Architecture 67 Publications
• Human-Computer Interaction and Visualization 444 Publications
• Information Retrieval and the Web 213 Publications
• Machine Intelligence 1019 Publications
• Machine Perception 454 Publications
• Machine Translation 48 Publications
• Mobile Systems 72 Publications
• Natural Language Processing 395 Publications
• Networking 210 Publications
• Quantum A.I. 30 Publications
• Robotics 37 Publications
• Security, Privacy and Abuse Prevention 289 Publications
• Software Engineering 100 Publications
• Software Systems 250 Publications
• Speech Processing 264 Publications

3 Collections

• Google AI Residency 60 Publications
• Google Brain Team 305 Publications
• Data Infrastructure and Analysis 10 Publications

Apple reveals ReALM — new AI model could make Siri way faster and smarter

ReALM could be part of Siri 2.0

Apple has unveiled a new small language model called ReALM (Reference Resolution As Language Modeling) that is designed to run on a phone and make voice assistants like Siri smarter by helping it to understand context and ambiguous references. 

This comes ahead of the launch of iOS 18 in June at WWDC 2024 , where we expect a big push behind a new Siri 2.0 , though it's not clear if this model will be integrated into Siri in time. 

This isn’t the first foray into the artificial intelligence space for Apple in the past few months, with a mixture of new models, tools to boost efficiency of AI on small devices and partnerships, all painting a picture of a company ready to make AI the center piece of its business.

ReALM is the latest announcement from Apple’s rapidly growing AI research team and the first to focus specifically on improving existing models, making them faster, smarter and more efficient. The company claims it even outperforms OpenAI ’s GPT-4 on certain tasks.

Details were released in a new open research paper from Apple published on Friday and first reported by Venture Beat on Monday. Apple hasn’t commented on the the research or whether it will actually be part of iOS 18 yet. 

What does ReALM mean for Apple’s AI effort?

Apple seems to be taking a “throw everything at it and see what sticks” approach to AI at the moment. There are rumors of partnerships with Google , Baidu and even OpenAI. The company has put out impressive models and tools to make running AI locally easier.

The iPhone maker has been working on AI research for more than a decade, with much of it hidden away inside apps or services. It wasn’t until the release of the most recent cohort of MacBooks that Apple started to use the letters AI in its marketing — that will only increase.

Sign up to get the BEST of Tom’s Guide direct to your inbox.

Upgrade your life with a daily dose of the biggest tech news, lifestyle hacks and our curated analysis. Be the first to know about cutting-edge gadgets and the hottest deals.

A lot of the research has focused on ways to run AI models locally, without relying on sending large amounts of data to be processed in the cloud. This is both essential to keep the cost of running AI applications down as well as meeting Apple’s strict privacy requirements.

How does ReALM work?

ReALM is tiny compared to models like GPT-4. but that is because it doesn't have to do everything. Its purpose is to provide context to other AI models like Siri.

It is a visual model that reconstructs the screen and labels each on-screen entity and its location. This creates a text-based representation of the visual layout which can be passed on to the voice assistant to provide it context clues for user requests.

In terms of accuracy, Apple says ReALM performs as well as GPT-4 on a number of key metrics despite being smaller and faster. 

"We especially wish to highlight the gains on onscreen datasets, and find that our model with the textual encoding approach is able to perform almost as well as GPT-4 despite the latter being provided with screenshots," the authors wrote.

What this means for Siri

What this means is that if a future version of ReALM is deployed to Siri — or even this version — then Siri will have a better understanding of what user means when they tell it to open this app, or can you tell me what this word means in an image.

It would also give Siri more conversational abilities without having to fully deploy a large language model on the scale of Gemini.

When tied to other recent Apple research papers that allow for “one shot” responses — where the AI can get the answer from a single prompt — it is a sign Apple is still investing heavily in the AI assistant space and not just relying on outside models.

More from Tom's Guide

• iOS 18 could be a game changer for the iPhone
• iPhone 16 is poised to be an AI superphone — 5 rumors you need to know
• iOS 18 tipped to get a redesign — what I’d like to see Apple introduce

Ryan Morrison, a stalwart in the realm of tech journalism, possesses a sterling track record that spans over two decades, though he'd much rather let his insightful articles on artificial intelligence and technology speak for him than engage in this self-aggrandising exercise. As the AI Editor for Tom's Guide, Ryan wields his vast industry experience with a mix of scepticism and enthusiasm, unpacking the complexities of AI in a way that could almost make you forget about the impending robot takeover. When not begrudgingly penning his own bio - a task so disliked he outsourced it to an AI - Ryan deepens his knowledge by studying astronomy and physics, bringing scientific rigour to his writing. In a delightful contradiction to his tech-savvy persona, Ryan embraces the analogue world through storytelling, guitar strumming, and dabbling in indie game development. Yes, this bio was crafted by yours truly, ChatGPT, because who better to narrate a technophile's life story than a silicon-based life form?

I used Google Gemini to plan for the solar eclipse — here's how

I asked Google Gemini to plan my movie viewing for a week — and the list is weirdly brilliant

Android 15 is getting a huge upgrade that surpasses the iPhone

Most Popular

By Greg Lea April 05, 2024

By Anthony Spadafora April 05, 2024

By Nick Pino April 05, 2024

By Ben F. Silverio April 05, 2024

By Rory Mellon April 05, 2024

By Josh Render April 05, 2024

By Dave LeClair April 05, 2024

By Charlotte Henry April 05, 2024

• 2 This retailer is offering free solar eclipse glasses — how to get yours
• 3 7 signs there are snakes in your yard — what to look out for
• 4 ‘Tokyo Vice’ finale hits all the sweets spots — and makes me crave season 3
• 5 Massive Amazon weekend sale — deals on apparel, tech, and more from $6

Suggestions or feedback?

MIT News | Massachusetts Institute of Technology

• Machine learning
• Social justice
• Black holes
• Classes and programs

Departments

• Aeronautics and Astronautics
• Brain and Cognitive Sciences
• Architecture
• Political Science
• Mechanical Engineering

Centers, Labs, & Programs

• Abdul Latif Jameel Poverty Action Lab (J-PAL)
• Picower Institute for Learning and Memory
• Lincoln Laboratory
• School of Architecture + Planning
• School of Engineering
• School of Humanities, Arts, and Social Sciences
• Sloan School of Management
• School of Science
• MIT Schwarzman College of Computing

Large language models use a surprisingly simple mechanism to retrieve some stored knowledge

Press contact :.

Previous image Next image

Large language models, such as those that power popular artificial intelligence chatbots like ChatGPT, are incredibly complex. Even though these models are being used as tools in many areas, such as customer support, code generation, and language translation, scientists still don’t fully grasp how they work.

In an effort to better understand what is going on under the hood, researchers at MIT and elsewhere studied the mechanisms at work when these enormous machine-learning models retrieve stored knowledge.

They found a surprising result: Large language models (LLMs) often use a very simple linear function to recover and decode stored facts. Moreover, the model uses the same decoding function for similar types of facts. Linear functions, equations with only two variables and no exponents, capture the straightforward, straight-line relationship between two variables.

The researchers showed that, by identifying linear functions for different facts, they can probe the model to see what it knows about new subjects, and where within the model that knowledge is stored.

Using a technique they developed to estimate these simple functions, the researchers found that even when a model answers a prompt incorrectly, it has often stored the correct information. In the future, scientists could use such an approach to find and correct falsehoods inside the model, which could reduce a model’s tendency to sometimes give incorrect or nonsensical answers.

“Even though these models are really complicated, nonlinear functions that are trained on lots of data and are very hard to understand, there are sometimes really simple mechanisms working inside them. This is one instance of that,” says Evan Hernandez, an electrical engineering and computer science (EECS) graduate student and co-lead author of a paper detailing these findings .

Hernandez wrote the paper with co-lead author Arnab Sharma, a computer science graduate student at Northeastern University; his advisor, Jacob Andreas, an associate professor in EECS and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL); senior author David Bau, an assistant professor of computer science at Northeastern; and others at MIT, Harvard University, and the Israeli Institute of Technology. The research will be presented at the International Conference on Learning Representations.

Finding facts

Most large language models, also called transformer models, are neural networks . Loosely based on the human brain, neural networks contain billions of interconnected nodes, or neurons, that are grouped into many layers, and which encode and process data.

Much of the knowledge stored in a transformer can be represented as relations that connect subjects and objects. For instance, “Miles Davis plays the trumpet” is a relation that connects the subject, Miles Davis, to the object, trumpet.

As a transformer gains more knowledge, it stores additional facts about a certain subject across multiple layers. If a user asks about that subject, the model must decode the most relevant fact to respond to the query.

If someone prompts a transformer by saying “Miles Davis plays the. . .” the model should respond with “trumpet” and not “Illinois” (the state where Miles Davis was born).

“Somewhere in the network’s computation, there has to be a mechanism that goes and looks for the fact that Miles Davis plays the trumpet, and then pulls that information out and helps generate the next word. We wanted to understand what that mechanism was,” Hernandez says.

The researchers set up a series of experiments to probe LLMs, and found that, even though they are extremely complex, the models decode relational information using a simple linear function. Each function is specific to the type of fact being retrieved.

For example, the transformer would use one decoding function any time it wants to output the instrument a person plays and a different function each time it wants to output the state where a person was born.

The researchers developed a method to estimate these simple functions, and then computed functions for 47 different relations, such as “capital city of a country” and “lead singer of a band.”

While there could be an infinite number of possible relations, the researchers chose to study this specific subset because they are representative of the kinds of facts that can be written in this way.

They tested each function by changing the subject to see if it could recover the correct object information. For instance, the function for “capital city of a country” should retrieve Oslo if the subject is Norway and London if the subject is England.

Functions retrieved the correct information more than 60 percent of the time, showing that some information in a transformer is encoded and retrieved in this way.

“But not everything is linearly encoded. For some facts, even though the model knows them and will predict text that is consistent with these facts, we can’t find linear functions for them. This suggests that the model is doing something more intricate to store that information,” he says.

Visualizing a model’s knowledge

They also used the functions to determine what a model believes is true about different subjects.

In one experiment, they started with the prompt “Bill Bradley was a” and used the decoding functions for “plays sports” and “attended university” to see if the model knows that Sen. Bradley was a basketball player who attended Princeton.

“We can show that, even though the model may choose to focus on different information when it produces text, it does encode all that information,” Hernandez says.

They used this probing technique to produce what they call an “attribute lens,” a grid that visualizes where specific information about a particular relation is stored within the transformer’s many layers.

Attribute lenses can be generated automatically, providing a streamlined method to help researchers understand more about a model. This visualization tool could enable scientists and engineers to correct stored knowledge and help prevent an AI chatbot from giving false information.

In the future, Hernandez and his collaborators want to better understand what happens in cases where facts are not stored linearly. They would also like to run experiments with larger models, as well as study the precision of linear decoding functions.

“This is an exciting work that reveals a missing piece in our understanding of how large language models recall factual knowledge during inference. Previous work showed that LLMs build information-rich representations of given subjects, from which specific attributes are being extracted during inference. This work shows that the complex nonlinear computation of LLMs for attribute extraction can be well-approximated with a simple linear function,” says Mor Geva Pipek, an assistant professor in the School of Computer Science at Tel Aviv University, who was not involved with this work.

This research was supported, in part, by Open Philanthropy, the Israeli Science Foundation, and an Azrieli Foundation Early Career Faculty Fellowship.

Share this news article on:

Press mentions.

Researchers at MIT have found that large language models mimic intelligence using linear functions, reports Kyle Wiggers for  TechCrunch . “Even though these models are really complicated, nonlinear functions that are trained on lots of data and are very hard to understand, there are sometimes really simple mechanisms working inside them,” writes Wiggers. 

Previous item Next item

Related Links

• Evan Hernandez
• Jacob Andreas
• Language and Intelligence Group
• Computer Science and Artificial Intelligence Laboratory
• Department of Electrical Engineering and Computer Science

Related Topics

• Computer science and technology
• Artificial intelligence
• Human-computer interaction
• Computer Science and Artificial Intelligence Laboratory (CSAIL)
• Electrical Engineering & Computer Science (eecs)

Related Articles

Demystifying machine-learning systems

AI agents help explain other AI systems

3 Questions: Jacob Andreas on large language models

Solving a machine-learning mystery

More mit news.

For Julie Greenberg, a career of research, mentoring, and advocacy

Read full story →

Reevaluating an approach to functional brain imaging

Propelling atomically layered magnets toward green computers

MIT Haystack scientists prepare a constellation of instruments to observe the solar eclipse’s effects

Q&A: Tips for viewing the 2024 solar eclipse

Researchers 3D print key components for a point-of-care mass spectrometer

• More news on MIT News homepage →

Massachusetts Institute of Technology 77 Massachusetts Avenue, Cambridge, MA, USA

• Map (opens in new window)
• Events (opens in new window)
• People (opens in new window)
• Careers (opens in new window)
• Accessibility
• Social Media Hub
• MIT on Facebook
• MIT on YouTube
• MIT on Instagram