Pornographic user query identification method and device for search engine

Abstract

The invention provides a pornographic user query identification method for a search engine. The method comprises the following steps of: training a classification model used for identifying pornographic user query based on the labeled user query set, and judging whether the user query to be identified is pornographic user query by utilizing the trained classification model, wherein the user query characteristics for training the classification model comprise the named entity and entity context subject characteristics of the user query and the probability that the user belongs to pornographic query and non-pornographic query. According to the method, the pornographic query is not required to be identified based on the query result of the search engine, and the pornographic query identification is rapidly and effectively realized by utilizing the named entity characteristics and linguistic model characteristics contained in the user query.

Description

technical field [0001] The invention relates to information retrieval and search engines, in particular to a method for identifying pornographic user queries in search engines. Background technique [0002] As the scale of the Internet continues to increase, search engines have become one of the important application systems on the Internet. How to accurately understand the query intent submitted by users to search engines is an important technical issue. Accurately understanding the user's query intent can help the search engine return better results or call the corresponding vertical search engine, thereby improving the user experience of the search engine. However, the information on the Internet is complicated, and the dross and the essence coexist, and the results returned by search engines often include some bad results, such as pornographic pages. Correctly identifying pornographic intent in user queries can help search engines take special action on pornographic us...

AI Technical Summary

Problems solved by technology

This type of method uses the website or page that the user clicks on the query result page to identify the query intent. This type of method assumes that the page that the user clicks meets the query intent, but when the user clicks the page and leaves quickly, it often means that the query intent does not. is satisfied, this behavior pattern is not reflected in the user click log

Another type of method uses the search results of the user query to reversely judge the intent of the user query. This method uses the topic distribution characteristics of the search result page to identify the distribution of the user query intent. The recognition is not accurate enough

Moreover, this method of reversely judging the user's query intent based on the search results has a high time complexity in identifying the query intent, which causes a large load on the search engine.

Images