Professor Liu Yan: Next generation search is blurred boundary

"Bit network message" although there are many expressions and statements about the next generation of search technology, but in the Information Retrieval Research Department Director, the national 863 "Chinese processing" key project overall group expert Liu Yan view, it is difficult to say what is the next generation, intelligent search is still in the fuzzy boundary, but some characteristics and technical trends have emerged.

Liu Yan, director of Information Retrieval Research Laboratory, Harbin

Personalized Retrieval

In the Internet search field has done such a test, when the search page logo masked, while in several large Internet search tools to type keywords to search, the results of Google's search results are most satisfying. In the field of Internet search technology, Google-represented keyword search is saturated enough.

"The big search based on keywords has developed almost," Liu Yan, a personalized search. At present, when people do keyword search, different people type the same keyword to get the results are exactly the same, personalized search can be based on the location of the searcher, the previous search history and other comprehensive factors, through tracking and retrieval tools to guess what the searcher want. For example, enter a keyword "Apple", some people want to look for Apple products, some people want to get some information about the fruit.

There is, of course, a controversy over the need for personalization, for example, when the same person is involved in "Sina", the result is the same, without personalization. However, personalization will also be the current Internet search business model to play a role in helping push, when the search technology enough personalized, for tracking search mature enough, personalized advertising will become

Very valuable. For example, a user who wants to buy a real estate, when the search technology guesses its true ideas, can be accurate to the real estate related information and advertising delivery to those in need.

Liu Yan that, at present, personalized search technology is not too difficult in the algorithm, but this kind of retrieval for the computer itself is very high requirements. With the current development of cloud computing and other related technologies, personalized search will also trend.

Question and answer system: to the simple

According to CNNIC's latest survey report, currently, users of the search results are not satisfied with the overall 35%,40% of the user will be the first time to type keywords based on the revision, 72% of users think the search results are very messy. In the vast amount of information on the Internet, the simple and clear search for what you want without any other interference, it also depends on the search technology to the searcher's idea of speculation.

Professor Liu Yan Analysis, at present, we see the search results are "reference" way, rather than the answer to you, this is the most existing search technology problems.

In the existing search technology, there is already such a way to "junk" information for the primary exclusion, such as "Baidu Know", when a person asked a question, after a person or more people to find their own answers to the information needed. On the other hand, the major manufacturers are also committed to achieve real "man-machine dialogue", with Powerset and Wolframalpha as the representative of the question-and-answer type of accurate search recently received widespread attention. The most important feature of this search is simplicity and clarity.

　Mobile search is not a simple PC application transplant

And with the gradual decline in mobile phone internet tariffs, 3G large-scale commercial, mobile terminal applications become more and more sophisticated, such as UCWeb will be enough to make the Web page beautiful and concise, many search vendors began to work with telecom operators and mobile terminal providers bundled with the mobile terminal to preinstall its search tools. In Microsoft's newly launched search tool Bing (Bing), Microsoft MSN Vice President Liuzhenyu said, has now launched mobile search with China Telecom cooperation.

However, in Professor Liu Yan's view, mobile search is not simply a simple transplant of a search application on a PC. "It (mobile search) has a specific search requirements, the requirements of special precision, and the location of the searcher is closely related to", while Liu Yan that mobile search applications are more life-class search.

Search for Community Business

Pay-per-click or bidding rankings, since the 1997 search field has emerged, major manufacturers have been using such a business model, which makes such as vertical search sites because the user group is too small and has been in the slit to survive. The innovation of search business model becomes the topic that the major manufacturers pay attention to at present.

On the one hand, search because of massive data mining makes it seem omnipotent, on the other hand, because search technology is ubiquitous and the impact of traditional areas to other commercial interest groups. Some of the world's biggest record companies have sued Baidu for illegal downloads in previous years, and the way the big labels are now spreading the internet is looking for a slice of the internet. So with the traditional record companies and search companies to cooperate to provide users with free authentic music. The ubiquitous tentacles of the internet make its business model break the traditional way and begin to alliances.

Liu Yan that, with the rapid rise of SNS, search and Internet Community resources integration will be likely to bring innovative business model. Make internet advertising more credible and credible, for example, through word of mouth from people who are familiar with each other on Facebook.

In addition, the search field has been exploring intelligent ways such as voice search, Liu Yan that voice search is mainly in speech recognition and speech synthesis, although the current speech synthesis technology has matured, but the semantic recognition will be because of noise and other interference is difficult to practical.