Introducing the Knowledge Graph: things, not strings

Cross-posted from the Official Google Blog

Search is a lot about discovery—the basic human need to learn and broaden your horizons. But searching still requires a lot of hard work by you, the user. So today I’m really excited to launch the Knowledge Graph, which will help you discover new information quickly and easily.

Take a query like [taj mahal]. For more than four decades, search has essentially been about matching keywords to queries. To a search engine the words [taj mahal] have been just that—two words.

But we all know that [taj mahal] has a much richer meaning. You might think of one of the world’s most beautiful monuments, or a Grammy Award-winning musician, or possibly even a casino in Atlantic City, NJ. Or, depending on when you last ate, the nearest Indian restaurant. It’s why we’ve been working on an intelligent model—in geek-speak, a “graph”—that understands real-world entities and their relationships to one another: things, not strings.

The Knowledge Graph enables you to search for things, people or places that Google knows about—landmarks, celebrities, cities, sports teams, buildings, geographical features, movies, celestial objects, works of art and more—and instantly get information that’s relevant to your query. This is a critical first step towards building the next generation of search, which taps into the collective intelligence of the web and understands the world a bit more like people do.

Google’s Knowledge Graph isn’t just rooted in public sources such as Freebase, Wikipedia and the CIA World Factbook. It’s also augmented at a much larger scale—because we’re focused on comprehensive breadth and depth. It currently contains more than 500 million objects, as well as more than 3.5 billion facts about and relationships between these different objects. And it’s tuned based on what people search for, and what we find out on the web.

The Knowledge Graph enhances Google Search in three main ways to start:

1. Find the right thing
Language can be ambiguous—do you mean Taj Mahal the monument, or Taj Mahal the musician? Now Google understands the difference, and can narrow your search results just to the one you mean—just click on one of the links to see that particular slice of results:

This is one way the Knowledge Graph makes Google Search more intelligent—your results are more relevant because we understand these entities, and the nuances in their meaning, the way you do.

2. Get the best summary
With the Knowledge Graph, Google can better understand your query, so we can summarize relevant content around that topic, including key facts you’re likely to need for that particular thing. For example, if you’re looking for Marie Curie, you’ll see when she was born and died, but you’ll also get details on her education and scientific discoveries:

How do we know which facts are most likely to be needed for each item? For that, we go back to our users and study in aggregate what they’ve been asking Google about each item. For example, people are interested in knowing what books Charles Dickens wrote, whereas they’re less interested in what books Frank Lloyd Wright wrote, and more in what buildings he designed.

The Knowledge Graph also helps us understand the relationships between things. Marie Curie is a person in the Knowledge Graph, and she had two children, one of whom also won a Nobel Prize, as well as a husband, Pierre Curie, who claimed a third Nobel Prize for the family. All of these are linked in our graph. It’s not just a catalog of objects; it also models all these inter-relationships. It’s the intelligence between these different entities that’s the key.

3. Go deeper and broader
Finally, the part that’s the most fun of all—the Knowledge Graph can help you make some unexpected discoveries. You might learn a new fact or new connection that prompts a whole new line of inquiry. Do you know where Matt Groening, the creator of the Simpsons (one of my all-time favorite shows), got the idea for Homer, Marge and Lisa’s names? It’s a bit of a surprise:

We’ve always believed that the perfect search engine should understand exactly what you mean and give you back exactly what you want. And we can now sometimes help answer your next question before you’ve asked it, because the facts we show are informed by what other people have searched for. For example, the information we show for Tom Cruise answers 37 percent of next queries that people ask about him. In fact, some of the most serendipitous discoveries I’ve made using the Knowledge Graph are through the magical “People also search for” feature. One my favorite books is The White Tiger, the debut novel by Aravind Adiga, which won the prestigious Man Booker Prize. Using the Knowledge Graph, I discovered three other books that had won the same prize and one that won the Pulitzer. I can tell you, this suggestion was spot on!

We’ve begun to gradually roll out this view of the Knowledge Graph to U.S. English users. It’s also going to be available on smartphones and tablets—read more about how we’ve tailored this to mobile devices. And watch our video (also available on our site about the Knowledge Graph) that gives a deeper dive into the details and technology, in the words of people who've worked on this project:

We hope this added intelligence will give you a more complete picture of your interest, provide smarter search results, and pique your curiosity on new topics. We’re proud of our first baby step—the Knowledge Graph—which will enable us to make search more intelligent, moving us closer to the "Star Trek computer" that I've always dreamt of building. Enjoy your lifelong journey of discovery, made easier by Google Search, so you can spend less time searching and more time doing what you love.

Posted by Amit Singhal, SVP, Engineering

The Knowledge Graph for mobile and tablet search

Have you ever had a question pop into your head at an unexpected time? Maybe when you’re talking with a friend over lunch, watching TV at home, or reading a magazine on the bus? On smartphones and tablets Google is great for these types of situations because it puts the information of the entire web at your fingertips. Today, we’re making it even faster and easier to get answers and explore no matter where you are, with the launch of the Knowledge Graph on desktop, smartphones, and tablets.

On wireless networks and on small screens, every page load and every pixel matters when it comes to speed and ease-of-use. So we strive for efficiency and try to make the most of touch-based interactions when integrating information from Knowledge Graph into our mobile and tablet search experiences.

For example, say this fall I’m heading to Chicago for a friend’s wedding, and I’ve heard I should check out Millennium Park while I’m in town. A quick search on Google brings up Knowledge Graph information embedded within the results. This initial peek shows what people are often interested in about Millennium Park.




Tapping or swiping on the content from the Knowledge Graph instantly shows me more useful information. I can see if there’s an event going on while I’m in town, and get some ideas for other Chicago attractions I might want to visit based on what other people have searched for on Google.

When searching on my tablet, I can swipe the rows of images to explore more related content.

Now let’s take another example. Say I’m searching for [andromeda], which could be the galaxy, the TV series, or the Swedish band. The Knowledge Graph distinguishes between each of these meanings and shows me an interactive ribbon at the top of the search results that I can swipe and tap to select just what I’m looking for. That means less typing.




These features are currently rolling out to most Android 2.2+ and iOS4+ devices. On Android, the feature is available through Google in the browser and the Quick Search Box. On iOS, the feature is available in the browser and will be coming soon to the Google Search App.


Posted by Junyoung Lee, Engineering Manager

Test your search skills with A Google a Day now on Google+

A Google a Day is the trivia game where searching for the answer on Google is not only allowed but encouraged. Since launching A Google a Day a year ago, millions of people have played and learned new trivia facts along with new ways to use Google search. Over the past year, you may have seen some questions written by our celebrity guest authors (Jane Goodall, Taylor Swift, and Vanessa Carlton, to name a few) and learned a search trick or two along the way.

One piece of feedback we’ve continually heard from people is that they want to be able to share their scores, challenge their friends, and answer more than one question a day. So today, we’re introducing a new and improved version of A Google a Day on Google+ that makes the trivia more collaborative and fun.

Now you’ll be able to challenge your friends to a test of trivia and search speed and figure out who’s the expert at discovering little known facts about literature, history, or science. Race to the top of the leaderboard and earn bragging rights to show off your master search skills. Through fast-paced competition and an exciting battle of the wits, we hope the new A Google a Day game triggers your imagination about all the types of questions you can ask Google.

Check out a sample question below:

Start playing A Google a Day now—Click here to access the new version of the game (you’ll be prompted to log into your Google+ account) or look for A Google a Day in the Games section of Google+. The clues are currently only in English, but anyone can attempt to solve the puzzles. And let us know what you think on our Google+ page (tag your posts with #agoogleaday), on Twitter at @agoogleaday, or at agoogleaday@google.com.

Posted by Dan Russell, User Experience Researcher

Search quality highlights: 52 changes for April

Update 6 May, 950am: We accidentally had one change included twice, "No freshness boost for low-quality content." We've removed the duplicate entry and updated the number of total launches from 53+ to 52+. - Ed.

We’ve had a zerg rush of 52+ launches this month in search. One of the big changes for me was our latest algorithm improvement to help you find more high-quality sites. But, that’s not all we’ve been up to. As you may recall, a couple months back we shared uncut video discussion of a spelling related change, and now that’s launched as well (see “More spell corrections for long queries”). Other highlights include changes in indexing, spelling, sitelinks, sports scores features and more. We even experimented with a couple more radical features, such as Really Advanced Search and Weather Control, but ultimately decided they were a little too foolish.

Here’s the real list for April:

• Categorize paginated documents. [launch codename "Xirtam3", project codename "CategorizePaginatedDocuments"] Sometimes, search results can be dominated by documents from a paginated series. This change helps surface more diverse results in such cases.
• More language-relevant navigational results. [launch codename "Raquel"] For navigational searches when the user types in a web address, such as [bol.com], we generally try to rank that web address at the top. However, this isn’t always the best answer. For example, bol.com is a Dutch page, but many users are actually searching in Portuguese and are looking for the Brazilian email service, http://www.bol.uol.com.br/. This change takes into account language to help return the most relevant navigational results.
• Country identification for webpages. [launch codename "sudoku"] Location is an important signal we use to surface content more relevant to a particular country. For a while we’ve had systems designed to detect when a website, subdomain, or directory is relevant to a set of countries. This change extends the granularity of those systems to the page level for sites that host user generated content, meaning that some pages on a particular site can be considered relevant to France, while others might be considered relevant to Spain.
• Anchors bug fix. [launch codename "Organochloride", project codename "Anchors"] This change fixed a bug related to our handling of anchors.
• More domain diversity. [launch codename "Horde", project codename "Domain Crowding"] Sometimes search returns too many results from the same domain. This change helps surface content from a more diverse set of domains.
• More local sites from organizations. [project codename "ImpOrgMap2"] This change makes it more likely you’ll find an organization website from your country (e.g. mexico.cnn.com for Mexico rather than cnn.com).
• Improvements to local navigational searches. [launch codename "onebar-l"] For searches that include location terms, e.g. [dunston mint seattle] or [Vaso Azzurro Restaurant 94043], we are more likely to rank the local navigational homepages in the top position, even in cases where the navigational page does not mention the location.
• Improvements to how search terms are scored in ranking. [launch codename "Bi02sw41"] One of the most fundamental signals used in search is whether and how your search terms appear on the pages you’re searching. This change improves the way those terms are scored.
• Disable salience in snippets. [launch codename "DSS", project codename "Snippets"] This change updates our system for generating snippets to keep it consistent with other infrastructure improvements. It also simplifies and increases consistency in the snippet generation process.
• More text from the beginning of the page in snippets. [launch codename "solar", project codename "Snippets"] This change makes it more likely we’ll show text from the beginning of a page in snippets when that text is particularly relevant.
• Smoother ranking changes for fresh results. [launch codename "sep", project codename "Freshness"] We want to help you find the freshest results, particularly for searches with important new web content, such as breaking news topics. We try to promote content that appears to be fresh. This change applies a more granular classifier, leading to more nuanced changes in ranking based on freshness.
• Improvement in a freshness signal. [launch codename "citron", project codename "Freshness"] This change is a minor improvement to one of the freshness signals which helps to better identify fresh documents.
• No freshness boost for low-quality content. [launch codename “NoRot”, project codename “Freshness”] We have modified a classifier we use to promote fresh content to exclude fresh content identified as particularly low-quality.
• Tweak to trigger behavior for Instant Previews. This change narrows the trigger area for Instant Previews so that you won’t see a preview until you hover and pause over the icon to the right of each search result. In the past the feature would trigger if you moused into a larger button area.
• Sunrise and sunset search feature internationalization. [project codename "sunrise-i18n"] We’ve internationalized the sunrise and sunset search feature to 33 new languages, so now you can more easily plan an evening jog before dusk or set your alarm clock to watch the sunrise with a friend. 
• Improvements to currency conversion search feature in Turkish. [launch codename "kur", project codename "kur"] We launched improvements to the currency conversion search feature in Turkish. Try searching for [dolar kuru], [euro ne kadar], or [avro kaç para].
• Improvements to news clustering for Serbian. [launch codename "serbian-5"] For news results, we generally try to cluster articles about the same story into groups. This change improves clustering in Serbian by better grouping articles written in Cyrillic and Latin. We also improved our use of “stemming” -- a technique that relies on the “stem” or root of a word.
• Better query interpretation. This launch helps us better interpret the likely intention of your search query as suggested by your last few searches.
• News universal results serving improvements. [launch codename "inhale"] This change streamlines the serving of news results on Google by shifting to a more unified system architecture.
• UI improvements for breaking news topics. [launch codename "Smoothie", project codename "Smoothie"] We’ve improved the user interface for news results when you’re searching for a breaking news topic. You’ll often see a large image thumbnail alongside two fresh news results.
• More comprehensive predictions for local queries. [project codename "Autocomplete"] This change improves the comprehensiveness of autocomplete predictions by expanding coverage for long-tail U.S. local search queries such as addresses or small businesses.
• Improvements to triggering of public data search feature. [launch codename "Plunge_Local", project codename "DIVE"] This launch improves triggering for the public data search feature, broadening the range of queries that will return helpful population and unemployment data.
• Adding Japanese and Korean to error page classifier. [launch codename "maniac4jars", project codename "Soft404"] We have signals designed to detect crypto 404 pages (also known as “soft 404s”), pages that return valid text to a browser, but the text only contains error messages, such as “Page not found.” It’s rare that a user will be looking for such a page, so it’s important we be able to detect them. This change extends a particular classifier to Japanese and Korean.
• More efficient generation of alternative titles. [launch codename "HalfMarathon"] We use a variety of signals to generate titles in search results. This change makes the process more efficient, saving tremendous CPU resources without degrading quality.
• More concise and/or informative titles. [launch codename "kebmo"] We look at a number of factors when deciding what to show for the title of a search result. This change means you’ll find more informative titles and/or more concise titles with the same information.
• Fewer bad spell corrections internationally. [launch codename "Potage", project codename "Spelling"] When you search for [mango tea], we don't want to show spelling predictions like “Did you mean 'mint tea'?” We have algorithms designed to prevent these “bad spell corrections” and this change internationalizes one of those algorithms.
• More spelling corrections globally and in more languages. [launch codename "pita", project codename "Autocomplete"] Sometimes autocomplete will correct your spelling before you’ve finished typing. We’ve been offering advanced spelling corrections in English, and recently we extended the comprehensiveness of this feature to cover more than 60 languages.
• More spell corrections for long queries. [launch codename "caterpillar_new", project codename "Spelling"] We rolled out a change making it more likely that your query will get a spell correction even if it’s longer than ten terms. You can watch uncut footage of when we decided to launch this from our past blog post.
• More comprehensive triggering of “showing results for” goes international. [launch codename "ifprdym", project codename "Spelling"] In some cases when you’ve misspelled a search, say [pnumatic], the results you find will actually be results for the corrected query, “pneumatic.” In the past, we haven’t always provided the explicit user interface to say, “Showing results for pneumatic” and the option to “Search instead for pnumatic.” We recently started showing the explicit “Showing results for” interface more often in these cases in English, and now we’re expanding that to new languages.
• “Did you mean” suppression goes international. [launch codename "idymsup", project codename "Spelling"] Sometimes the “Did you mean?” spelling feature predicts spelling corrections that are accurate, but wouldn’t actually be helpful if clicked. For example, the results for the predicted correction of your search may be nearly identical to the results for your original search. In these cases, inviting you to refine your search isn’t helpful. This change first checks a spell prediction to see if it’s useful before presenting it to the user. This algorithm was already rolled out in English, but now we’ve expanded to new languages. 
• Spelling model refresh and quality improvements. We’ve refreshed spelling models and launched quality improvements in 27 languages.
• Fewer autocomplete predictions leading to low-quality results. [launch codename "Queens5", project codename "Autocomplete"] We’ve rolled out a change designed to show fewer autocomplete predictions leading to low-quality results.
• Improvements to SafeSearch for videos and images. [project codename "SafeSearch"] We’ve made improvements to our SafeSearch signals in videos and images mode, making it less likely you’ll see adult content when you aren’t looking for it.
• Improved SafeSearch models. [launch codename "Squeezie", project codename "SafeSearch"] This change improves our classifier used to categorize pages for SafeSearch in 40+ languages.
• Improvements to SafeSearch signals in Russian. [project codename "SafeSearch"] This change makes it less likely that you’ll see adult content in Russian when you aren’t looking for it.
• Increase base index size by 15%. [project codename "Indexing"] The base search index is our main index for serving search results and every query that comes into Google is matched against this index. This change increases the number of documents served by that index by 15%. *Note: We’re constantly tuning the size of our different indexes and changes may not always appear in these blog posts.
• New index tier. [launch codename "cantina", project codename "Indexing"] We keep our index in “tiers” where different documents are indexed at different rates depending on how relevant they are likely to be to users. This month we introduced an additional indexing tier to support continued comprehensiveness in search results.
• Backend improvements in serving. [launch codename "Hedges", project codename "Benson"] We’ve rolled out some improvements to our serving systems making them less computationally expensive and massively simplifying code.
• "Sub-sitelinks" in expanded sitelinks. [launch codename "thanksgiving"] This improvement digs deeper into megasitelinks by showing sub-sitelinks instead of the normal snippet.
• Better ranking of expanded sitelinks. [project codename "Megasitelinks"] This change improves the ranking of megasitelinks by providing a minimum score for the sitelink based on a score for the same URL used in general ranking.
• Sitelinks data refresh. [launch codename "Saralee-76"] Sitelinks (the links that appear beneath some search results and link deeper into the site) are generated in part by an offline process that analyzes site structure and other data to determine the most relevant links to show users. We’ve recently updated the data through our offline process. These updates happen frequently (on the order of weeks).
• Less snippet duplication in expanded sitelinks. [project codename "Megasitelinks"] We’ve adopted a new technique to reduce duplication in the snippets of expanded sitelinks.
• Movie showtimes search feature for mobile in China, Korea and Japan. We’ve expanded our movie showtimes feature for mobile to China, Korea and Japan.
• MLB search feature. [launch codename "BallFour", project codename "Live Results"] As the MLB season began, we rolled out a new MLB search feature. Try searching for [sf giants score] or [mlb scores].
• Spanish football (La Liga) search feature. This feature provides scores and information about teams playing in La Liga. Try searching for [barcelona fc] or [la liga].
• Formula 1 racing search feature. [launch codename "CheckeredFlag"] This month we introduced a new search feature to help you find Formula 1 leaderboards and results. Try searching [formula 1] or [mark webber].
• Tweaks to NHL search feature. We’ve improved the NHL search feature so it’s more likely to appear when relevant. Try searching for [nhl scores] or [capitals score].
• Keyword stuffing classifier improvement. [project codename "Spam"] We have classifiers designed to detect when a website is keyword stuffing. This change made the keyword stuffing classifier better. 
• More authoritative results. We’ve tweaked a signal we use to surface more authoritative content. 
• Better HTML5 resource caching for mobile. We’ve improved caching of different components of the search results page, dramatically reducing latency in a number of cases.

And here are some other changes we’ve blogged about since last time:

• Updates to rich snippets
• Another step to reward high-quality sites 

Posted by Matt Cutts, Distinguished Engineer

Helping students become savvy searchers with our new Search Education hub

The other day, some students and I were looking at a picture on a website, identified only as a page from an old encyclopedia. We wanted to find the original source, and tried dragging the picture into the Google Images search box, triggering the Search by Image tool. Image Search matched the picture with others copies found on the web: not only did we find the original source (an encyclopedia from 1826!)--we found a whole new world of information.

The students’ eyes lit up when they understood there was a technological answer to a problem they commonly encounter, and one which helped them increase the quality of their work as well. The technology gave them the opportunity to be detectives and ferret out the “truth” behind an image.

In the interest of creating more of these magical moments for students, we’re pleased to announce our new Search Education website, bringing educators the tools they need to help students become savvy searchers and independent learners.



Learning to tap the full potential of Google Search empowers students in both their academic and daily pursuits. Search education provides the technical tools and critical thinking skills crucial to preparing today’s students to be technologically self-reliant, independent learners.

For the past several years, the Google Search Education team has offered classes to help educators around the world discover Google Search tools that allow for great precision and depth in academic research. Knowing how to search has so many benefits, but we know teachers need materials in order to teach these skills. That’s why we created our new Search Education hub, complete with lesson plans based upon the Common Core State Standards and “A Google a Day” search challenges, as well as trainings you can watch.

We look forward to hearing about your experiences teaching search techniques in the classroom, and receiving feedback that will help us continually improve Search Education offerings. Feel free to drop us a line at search-educators@google.com.


Posted by Tasha Bergson-Michelson, Google Search Educator

Another step to reward high-quality sites

(Cross-posted on the Webmaster Central Blog)

Google has said before that search engine optimization, or SEO, can be positive and constructive—and we're not the only ones. Effective search engine optimization can make a site more crawlable and make individual pages more accessible and easier to find. Search engine optimization includes things as simple as keyword research to ensure that the right words are on the page, not just industry jargon that normal people will never type.

“White hat” search engine optimizers often improve the usability of a site, help create great content, or make sites faster, which is good for both users and search engines. Good search engine optimization can also mean good marketing: thinking about creative ways to make a site more compelling, which can help with search engines as well as social media. The net result of making a great site is often greater awareness of that site on the web, which can translate into more people linking to or visiting a site.

The opposite of “white hat” SEO is something called “black hat webspam” (we say “webspam” to distinguish it from email spam). In the pursuit of higher rankings or traffic, a few sites use techniques that don’t benefit users, where the intent is to look for shortcuts or loopholes that would rank pages higher than they deserve to be ranked. We see all sorts of webspam techniques every day, from keyword stuffing to link schemes that attempt to propel sites higher in rankings.

The goal of many of our ranking changes is to help searchers find sites that provide a great user experience and fulfill their information needs. We also want the “good guys” making great sites for users, not just algorithms, to see their effort rewarded. To that end we’ve launched Panda changes that successfully returned higher-quality sites in search results. And earlier this year we launched a page layout algorithm that reduces rankings for sites that don’t make much content available “above the fold.”

In the next few days, we’re launching an important algorithm change targeted at webspam. The change will decrease rankings for sites that we believe are violating Google’s existing quality guidelines. We’ve always targeted webspam in our rankings, and this algorithm represents another improvement in our efforts to reduce webspam and promote high quality content. While we can't divulge specific signals because we don't want to give people a way to game our search results and worsen the experience for users, our advice for webmasters is to focus on creating high quality sites that create a good user experience and employ white hat SEO methods instead of engaging in aggressive webspam tactics.

Here’s an example of a webspam tactic like keyword stuffing taken from a site that will be affected by this change:

Of course, most sites affected by this change aren’t so blatant. Here’s an example of a site with unusual linking patterns that is also affected by this change. Notice that if you try to read the text aloud you’ll discover that the outgoing links are completely unrelated to the actual content, and in fact the page text has been “spun” beyond recognition:

Sites affected by this change might not be easily recognizable as spamming without deep analysis or expertise, but the common thread is that these sites are doing much more than white hat SEO; we believe they are engaging in webspam tactics to manipulate search engine rankings.

The change will go live for all languages at the same time. For context, the initial Panda change affected about 12% of queries to a significant degree; this algorithm affects about 3.1% of queries in English to a degree that a regular user might notice. The change affects roughly 3% of queries in languages such as German, Chinese, and Arabic, but the impact is higher in more heavily-spammed languages. For example, 5% of Polish queries change to a degree that a regular user might notice.

We want people doing white hat search engine optimization (or even no search engine optimization at all) to be free to focus on creating amazing, compelling web sites. As always, we’ll keep our ears open for feedback on ways to iterate and improve our ranking algorithms toward that goal.

Posted by Matt Cutts, Distinguished Engineer

Search quality highlights: 50 changes for March

Here’s our latest installment of search quality highlights, with another 50 changes to report for March. We’re starting to get into a groove with these posts, so we’re getting more and more comprehensive as the months go by. New for this month, we’ve published uncut video from our search quality meeting, which gives a great flavor for how these decisions get made.

Here’s the list for March:

• Autocomplete with math symbols. [launch codename "Blackboard", project codename "Suggest"] When we process queries to return predictions in autocomplete, we generally normalize them to match more relevant predictions in our database. This change incorporates several characters that were previously normalized: “+”, “-”, “*”, “/”, “^”, “(“, “)”, and “=”. This should make it easier to search for popular equations, for example [e = mc2] or [y = mx+b].
• Improvements to handling of symbols for indexing. [launch codename "Deep Maroon"] We generally ignore punctuation symbols in queries. Based on analysis of our query stream, we’ve now started to index the following heavily used symbols: “%”, “$”, “\”, “.”, “@”, “#”, and “+”. We’ll continue to index more symbols as usage warrants.
• Better scoring of news groupings. [launch codename "avenger_2"] News results on Google are organized into groups that are about the same story. We have scoring systems to determine the ordering of these groups for a given query. This subtle change slightly improves our scoring system, leading to better ranking of news clusters.
• Sitelinks data refresh. [launch codename "Saralee-76"] Sitelinks (the links that appear beneath some search results and link deeper into the respective site) are generated in part by an offline process that analyzes site structure and other data to determine the most relevant links to show users. We’ve recently updated the data through our offline process. These updates happen frequently (on the order of weeks).
• Improvements to autocomplete backends, coverage. [launch codename "sovereign", project codename "Suggest"] We’ve consolidated systems and reduced the number of backend calls required to prepare autocomplete predictions for your query. The result is more efficient CPU usage and more comprehensive predictions.
• Better handling of password changes. Our general approach is that when you change passwords, you’ll be signed out from your account on all machines. This change ensures that changing your password more consistently signs your account out of Search, everywhere.
• Better indexing of profile pages. [launch codename "Prof-2"] This change improves the comprehensiveness of public profile pages in our index from more than two-hundred social sites.
• UI refresh for News Universal. [launch codename "Cosmos Newsy", project codename "Cosmos"] We’ve refreshed the design of News Universal results by providing more results from the top cluster, unifying the UI treatment of clusters of different sizes, adding a larger font for the top article, adding larger images (from licensed sources), and adding author information.
• Improvements to results for navigational queries. [launch codename "IceMan5"] A “navigational query” is a search where it looks like the user is looking to navigate to a particular website, such as [New York Times] or [wikipedia.org]. While these searches may seem straightforward, there are still challenges to serving the best results. For example, what if the user doesn’t actually know the right URL? What if the URL they’re searching for seems to be a parked domain (with no content)? This change improves results for this kind of search.
• High-quality sites algorithm data update and freshness improvements. [launch codename “mm”, project codename "Panda"] Like many of the changes we make, aspects of our high-quality sites algorithm depend on processing that’s done offline and pushed on a periodic cycle. In the past month, we’ve pushed updated data for “Panda,” as we mentioned in a recent tweet. We’ve also made improvements to keep our database fresher overall.
• Live results for UEFA Champions League and KHL. We’ve added live-updating snippets in our search results for the KHL (Russian Hockey League) and UEFA Champions League, including scores and schedules. Now you can find live results from a variety of sports leagues, including the NFL, NBA, NHL and others.
• Tennis search feature. [launch codename "DoubleFault"] We’ve introduced a new search feature to provide realtime tennis scores at the top of the search results page. Try [maria sharapova] or [sony ericsson open].
• More relevant image search results. [launch codename "Lice"] This change tunes signals we use related to landing page quality for images. This makes it more likely that you’ll find highly relevant images, even if those images are on pages that are lower quality.
• Fresher image predictions in all languages. [launch codename "imagine2", project codename "Suggest"] We recently rolled out a change to surface more relevant image search predictions in autocomplete in English. This improvement extends the update to all languages.
• SafeSearch algorithm tuning. [launch codenames "Fiorentini", “SuperDyn”; project codename "SafeSearch"] This month we rolled out a couple of changes to our SafeSearch algorithm. We’ve updated our classifier to make it smarter and more precise, and we’ve found new ways to make adult content less likely to appear when a user isn't looking for it
• Tweaks to handling of anchor text. [launch codename "PC"] This month we turned off a classifier related to anchor text (the visible text appearing in links). Our experimental data suggested that other methods of anchor processing had greater success, so turning off this component made our scoring cleaner and more robust.
• Simplification to Images Universal codebase. [launch codename "Galactic Center"] We’ve made some improvements to simplify our codebase for Images Universal and to better utilize improvements in our general web ranking to also provide better image results.
• Better application ranking and UI on mobile. When you search for apps on your phone, you’ll now see richer results with app icons, star ratings, prices, and download buttons arranged to fit well on smaller screens. You’ll also see more relevant ranking of mobile applications based on your device platform, for example Android or iOS.
• Improvements to freshness in Video Universal. [launch codename "graphite", project codename "Freshness"] We’ve improved the freshness of video results to better detect stale videos and return fresh content.
• Fewer undesired synonyms. [project codename "Synonyms"] When you search on Google, we often identify other search terms that might have the same meaning as what you entered in the box (synonyms) and surface results for those terms as well when it might be helpful. This month we tweaked a classifier to prevent unhelpful synonyms from being introduced as content in the results set.
• Better handling of queries with both navigational and local intent. [launch codename "ShieldsUp"] Some queries have both local intent and are very navigational (directed towards a particular website). This change improves the balance of results we show, and helps ensure you’ll find highly relevant navigational results or local results towards the top of the page as appropriate for your query.
• Improvements to freshness. [launch codename "Abacus", project codename "Freshness"] We launched an improvement to freshness late last year that was very helpful, but it cost significant machine resources. At the time we decided to roll out the change only for news-related traffic. This month we rolled it out for all queries.
• Improvements to processing for detection of site quality. [launch codename "Curlup"] We’ve made some improvements to a longstanding system we have to detect site quality. This improvement allows us to get greater confidence in our classifications.
• Better interpretation and use of anchor text. We’ve improved systems we use to interpret and use anchor text, and determine how relevant a given anchor might be for a given query and website.
• Better local results and sources in Google News. [launch codename "barefoot", project codename "news search"] We’re deprecating a signal we had to help people find content from their local country, and we’re building similar logic into other signals we use. The result is more locally relevant Google News results and higher quality sources.
• Deprecating signal related to ranking in a news cluster. [launch codename "decaffeination", project codename "news search”] We’re deprecating a signal that’s no longer improving relevance in Google News. The signal was originally developed to help people find higher quality articles on Google News. (Note: Despite the launch codename, this project has nothing to do with Caffeine, our update to indexing in 2010).
• Fewer “sibling” synonyms. [launch codename "Gemini", project codename "Synonyms"] One of the main signals we look at to identify synonyms is context. For example, if the word “cat” often appears next to the term “pet” and “furry,” and so does the word “kitten”, our algorithms may guess that “cat” and “kitten” have similar meanings. The problem is that sometimes this method will introduce “synonyms” that actually are different entities in the same category. To continue the example, dogs are also “furry pets” -- so sometimes “dog” may be incorrectly introduced as a synonym for “cat”. We’ve been working for some time to appropriately ferret out these “sibling” synonyms, and our latest system is more maintainable, updatable, debuggable, and extensible to other systems.
• Better synonym accuracy and performance. [project codename "Synonyms"] We’ve made further improvements to our synonyms system by eliminating duplicate logic. We’ve also found ways to more accurately identify appropriate synonyms in cases where there are multiple synonym candidates with different contexts.
• Retrieval system tuning. [launch codename "emonga", project codename "Optionalization"] We’ve improved systems that identify terms in a query which are not necessarily required to retrieve relevant documents. This will make results more faithful to the original query.
• Less aggressive synonyms. [launch codename "zilong", project codename "Synonyms"] We’ve heard feedback from users that sometimes our algorithms are too aggressive at incorporating search results for other terms. The underlying cause is often our synonym system, which will include results for other terms in many cases. This change makes our synonym system less aggressive in the way it incorporates results for other query terms, putting greater weight on the original user query.
• Update to systems relying on geographic data. [launch codename "Maestro, Maitre"] We have a number of signals that rely on geographic data (similar to the data we surface in Google Earth and Maps). This change updates some of the geographic data we’re using.
• Improvements to name detection. [launch codename "edge", project codename "NameDetector"] We’ve improved a system for detecting names, particularly for celebrity names.
• Updates to personalization signals. [project codename "PSearch"] This change updates signals used to personalize search results.
• Improvements to Image Search relevance. [launch codename "sib"] We’ve updated signals to better promote reasonably sized images on high-quality landing pages.
• Remove deprecated signal from site relevance signals. [launch codename "Freedom"] We’ve removed a deprecated product-focused signal from a site-understanding algorithm.
• More precise detection of old pages. [launch codename "oldn23", project codename “Freshness"] This change improves detection of stale pages in our index by relying on more relevant signals. As a result, fewer stale pages are shown to users.
• Tweaks to language detection in autocomplete. [launch codename “Dejavu”, project codename "Suggest"] In general, autocomplete relies on the display language to determine what language predictions to show. For most languages, we also try to detect the user query language by analyzing the script, and this change extends that behavior to Chinese (Simplified and Traditional), Japanese and Korean. The net effect is that when users forget to turn off their IMEs, they’ll still get English predictions if they start typing English terms.
• Improvements in date detection for blog/forum pages. [launch codename "fibyen", project codename "Dates"] This change improves the algorithm that determines dates for blog and forum pages.
• More predictions in autocomplete by live rewriting of query prefixes. [launch codename "Lombart", project codename "Suggest”] In this change we’re rewriting partial queries on the fly to retrieve more potential matching predictions for the user query. We use synonyms and other features to get the best overall match. Rewritten prefixes can include term re-orderings, term additions, term removals and more.
• Expanded sitelinks on mobile. We’ve launched our expanded sitelinks feature for mobile browsers, providing better organization and presentation of sitelinks in search results.
• More accurate short answers. [project codename “Porky Pig”] We’ve updated the sources behind our short answers feature to rely on data from Freebase. This improves accuracy and makes it easier to fix bugs.
• Migration of video advanced search backends. We’ve migrated some backends used in video advanced search to our main search infrastructure.
• +1 button in search for more countries and domains. This month we’ve internationalized the +1 button on the search results page to additional languages and domains. The +1 button in search makes it easy to share recommendations with the world right from your search results. As we said in our initial blog post, the beauty of +1’s is their relevance—you get the right recommendations (because they come from people who matter to you), at the right time (when you are actually looking for information about that topic) and in the right format (your search results).
• Local result UI refresh on tablet. We’ve updated the user interface of local results on tablets to make them more compact and easier to scan.

And here are a few other changes we’ve blogged about since last time:

• Flights to worldwide destinations
• Redesigned Search App for Windows 7.5 phones
• SSL search around the globe
• “Recent” feature on mobile
• Full-page themes in iGoogle
• March Madness NCAA search feature
• 3D graphing calculator

Posted by Johanna Wright, Director of Product Management