The current lack of personalized search

Intermediary transaction SEO diagnosis Taobao guest Cloud host technology Hall

Generally speaking, the Web personalized information recommendation service used in search engine can draw on the Web personalized information recommendation service in traditional e-commerce field. However, the search engine with similar functions as the personalized recommendation system in the modern electronic commerce field has received little attention. In other words, modern search engines can not provide a wide range of personalized search results, the same query for different users of the search results are always the same, it is not related to the user submitting the query. Therefore, the search engine often returns a large amount of useless information for a particular user, because it ignores the user's personalized requirement characteristics during the retrieval phase. The main cause of this phenomenon is that modern search engine faces some problems similar to the traditional personalized information recommendation system, as follows.

First of all, user needs are difficult to express effectively. This is mainly two reasons: on the one hand, because the general non-professional users are lack of demand expression training, so can not effectively understand and express their information needs, the result is that the user's subjective understanding is often not clear. Figuratively speaking, the phenomenon is that "the user cannot describe what he is looking for unless he sees what he is looking for". On the other hand, this also comes from the system can not correctly obtain the relevant user personalized information, this is mainly because the network information retrieval system usually does not have the initiative to obtain personalized features of the user, but also does not require users must submit personalized information to use the restrictions. The final result of the above two aspects will result in the system cannot obtain the user personalization characteristic information effectively the phenomenon.

Secondly, there is a contradiction between the accuracy of retrieval results and the rapidity of retrieval. When dealing with the massive data of the search engine, many traditional personalized recommendation techniques often produce serious performance problems, which are mainly applicable to the traditional Small business web site personalization algorithms and technology often lack of good scalability. Of course, people have also proposed some solutions, such as dimensionality reduction, clustering analysis and Bayesian network, although to some extent can solve the problem of scaling, however, these technologies are often in the offline phase extracted from the original data pattern information, and in the online phase of the use of these patterns to get the recommended collection, Therefore, although these methods can reduce the processing overhead, it often produces inaccurate recommendation results, while the complexity of online computing increases with the increase of the pattern.

Finally, the level of retrieval intelligence of modern search engine is still low. Because the search engine can not really understand the semantics of the content of the Web page, the simple use of Word matching and statistical analysis will, to a certain extent, cause an inevitable error in understanding. It should be said that even the ability to use some intelligent semantic analysis, but also because of the cost of processing time and space, so in the commercial search engine system is not widely popular.

Author: Hangzhou SI billion Network Technology Co., Ltd.

Original load: http://www.seo.com.cn

Copyright Notice: Original works, allow reprint, reprint, please be sure to hyperlink form to indicate the original source of the article, the author's information and the

Statement。 Otherwise, legal liability will be held.