Search result diversification ordering method based on hierarchical structure subtopic

Abstract

The invention discloses a search result diversification ordering method based on a hierarchical structure subtopic. The method comprises the following steps: 1) defining an expression method of a hierarchical structure dendric subtopic of a search word; 2) estimating the correlation between the hierarchical structure subtopic and a searched document; and 3) establishing a search result diversification model based on the hierarchical structure subtopic of the search word; wherein step 3) is realized by any of two ordering methods: a) performing diversification ordering on documents according to a hierarchical structure topic novelty model; and b) performing diversification ordering on documents according to a hierarchical structure topic proportional model. The invention defines a searched hierarchical structure subtopic and a calculating method of correlation between multilayer subtopics and searched documents, and proposes a search result diversification algorithm based on the hierarchical structure subtopic, wherein a real user intention can be matched more accurately by flexibly using the subtopics of different granularities, so that the diversification of search results is improved.

Description

technical field [0001] The invention relates to a method for diversifying and sorting search results based on hierarchical structure subtopics. Background technique [0002] Internet information has covered people's daily life more and more comprehensively, and users have gradually become accustomed to relying on search engines to find the information they need. A large number of studies have shown that a considerable part of the queries submitted to search engines are short text queries. Due to the small amount of information, these short text queries are usually ambiguous or have multiple meanings when interpreting user intent. For common ambiguous queries, for example, when searching for "Apple", some users may be looking for information about the famous Apple company, while others are concerned about information about fruits and apples; when searching for "National People's Congress", some users may Maybe they are looking for information about Renmin University of Chin...

AI Technical Summary

Problems solved by technology

However, since the subtopics in the diversification algorithm are automatically generated according to the query, it is difficult to perfectly match the real user intent

However, the current diversification method mainly uses subtopics in the form of a list, and it is difficult to find subtopics with appropriate granularity that can perfectly match the real user intent.

Images