technologies behind google ranking
In my previous post, I introduced the philosophies behind Google ranking. As part of our effort to discuss search quality, I want to tell you more about the technologies behind our ranking. The core technology in our ranking system comes from the academic field of Information Retrieval (IR). The IR community has studied search for almost 50 years. It uses statistical signals of word salience, like word frequency, to rank pages. (See "Modern Information Retrieval: A Brief Overview" for a quick overview of IR technology.) IR gave us a solid foundation, and we have built a tremendous system on top using links, page structure, and many other such innovations.
Search in the last decade has moved from give me what I said to give me what I want. User expectations from search have rightly increased. We work hard to fulfill the expectations of each and every user, and to do that we need to better understand the pages, the queries, and our users. Over the last decade we have pushed the technologies for understanding these three components (of the search process) to completely new dimensions.
When we talk about queries at Google, we use square brackets [ ] to mark the beginning and end of queries (see "How to write queries" by Matt Cutts), a notation I will use throughout this post. (Pages and search results change frequently, so in time, some examples used here may not behave as explained.)
- Understanding pages: Over years we have invested heavily in our crawl and indexing system. As a result we have a very large and very fresh index. In addition to size and freshness, we have improved our index in other ways. One of the key technologies we have developed to understand pages is associating important concepts to a page even when they are not obvious on the page. We find the official homepage for Sprovieri Gallery in London for the Italian query [galleria sprovieri londra], even though the official page does not have either London or Londra on it. In the U.S., a user searching for [cool tech pc vancouver, wa] finds the homepage www.cooltechpc.com even though the page does not mention anywhere that they are in Vancouver, WA. Other technologies we have developed include distinctions between important and less important words in the page and the freshness of the information on the page.
- Understanding queries: It is critical that we understand what our users are looking for (beyond just the few words in their query). We have made several notable advances in this area including a best-in-class spelling suggestion system, an advanced synonyms system, and a very strong concept analysis system.
Most users have used our spelling suggestion system at one time or another. It knows that someone searching for [kofee annan] is really searching for Mr. Kofi Annan, and is prompted: Did you mean: kofi annan; whereas someone searching for [kofee beans] is actually looking for coffee beans. Doing this internationally with very high accuracy is hard, and we do it well.
Synonyms are the foundation of our query understanding work. This is one of the hardest problems we are solving at Google. Though sometimes obvious to humans, it is an unsolved problem in automatic language processing. As a user, I don't want to think too much about what words I should use in my queries. Often I don't even know what the right words are. This is where our synonyms system comes into action. Our synonyms system can do sophisticated query modifications, e.g., it knows that the word 'Dr' in the query [Dr Zhivago] stands for Doctor whereas in [Rodeo Dr] it means Drive. A user looking for [back bumper repair] gets results about rear bumper repair. For [Ramstein ab], we automatically look for Ramstein Air Base; for the query query [b&b ab] we search for Bed and Breakfasts in Alberta, Canada. We have developed this level of query understanding for almost one hundred different languages, which is what I am truly proud of.
Another technology we use in our ranking system is concept identification. Identifying critical concepts in the query allows us to return much more relevant results. For example, our algorithms understand that in the query [new york times square church] the user is looking for the well-known church in Times Square and not for articles from the New York Times. We don't just stop at identifying concepts; we further enhance the query with the right concepts when, for instance, someone looking for [PC and its impact on people] is in fact looking for impact of computers on society, or someone who searches for [rainforest instructional activities for vocabulary] is really looking for rain forest lesson plans. Our query analysis algorithms have many such state-of-the-art techniques built into them, and once again, we do this internationally in almost every language we serve.
- Understanding users: Our work on interpreting user intent is aimed at returning results people really want, not just what they said in their query. This work starts with a world class localization system, and adds to it our advanced personalization technology, and several other great strides we have made in interpreting user intent, e.g. Universal Search.
Our clear focus on "best locally relevant results served globally" is reflected in our work on localization. The same query typed in multiple countries may deserve completely different results. A user looking for [bank] in the US should get American banks, whereas a user in the UK is either looking for the Bank Fashion line or for British financial institutions. The results for this query should return local financial institutions in other English speaking countries like Australia, Canada, New Zealand, South Africa. The fun really starts when this query is typed in non-English-speaking countries like Egypt, Israel, Japan, Russia, Saudi Arabia, Switzerland. Likewise the query [football] refers to entirely different sports in Australia, the UK, and the US. These examples mostly show how we get the localized version of the same concept correctly (financial institution, sport, etc.). However, the same query can mean entirely different things in different countries. For example, [Côte d'Or] is a geographic region in France - but it is a large chocolate manufacturer in neighboring French-speaking Belgium; and yes, we get that right too :-).
Personalization is another strong feature in our search system which tailors search results to individual users. Users who are logged-in while searching and have signed up for Web History get results that are more relevant for them than the general Google results. For example, someone who does a lot football-related searches might get more football related results for [giants], while other users might get results related to the baseball team. Similarly, if you tend to prefer results from a particular shopping site, you will be more likely to get results from that site when you search for products. Our evaluation shows that users who get personalized results find them to be more relevant than non-personalized results.
Another case of user intent can be observed for the query [chevrolet magnum]. Magnum is actually made by Dodge and not Chevrolet. So we present the results for Dodge Magnum with the prompt See results for: dodge magnum in our result set.
Our work on Universal Search is another example of how we interpret user intent to give them what they (sometimes) really want. Someone searching for [bangalore] not only gets the important web pages, they also get a map, a video showing street life, traffic, etc. in Bangalore -- watching this video I almost feel I am there :-) -- and at the time of writing there is relevant news and relevant blogs about Bangalore.
Finally let me briefly mention the latest advance we have made in search: Cross Language Information Retrieval (CLIR). CLIR allows users to first discover information that is not in their language, and then using Google's translation technology, we make this information accessible. I call this advance: give me what I want in any language. A user looking for Tony Blair's biography in Russia who types the query in Russian [Тони Блэр биография] is prompted at the bottom of our results to search the English web with:
Similarly a user searching for Disney movie songs in Egypt with the query [أغاني أفلام ديزني] is prompted to search the English web. We are very excited about CLIR as it truly brings us closer to our mission to organize the world's information and make it universally accessible and useful.I hope my two posts about Google ranking have made it clear that we live and breathe search, and we are more passionate than ever about it. Our fervor for serving all our users worldwide is unprecedented. We pride ourselves in running a very good ranking system, and are working incredibly hard every day to make it even better.
I could go on and on showing examples of state-of-the-art technology that we have developed to make our ranking system as good as it is, but the fact is that search is nowhere close to being a solved problem. Many queries still don't get satisfactory results from Google, and each such query is an opportunity to improve our ranking system. I am confident that with numerous techniques under development in our group, we will make large improvements to our ranking algorithms in the near future.
Subscribe to:
Post Comments (Atom)
0 comments:
Post a Comment