Semantic search method, system and computer-readable storage medium

Abstract

The present invention provides a semantic search method, system and computer-readable storage medium, the method comprising: obtaining user-topic distribution, topic-word distribution, topic-topic tag distribution, topic by solving a pre-built social network multi-feature topic model ‑Time distribution; get the text of the item to be searched, user information, publication time and hashtag, get the topic matrix from the user information of the item to be searched and corresponding distribution, from the matrix, text and corresponding distribution, topic label and corresponding distribution, publication time and Corresponding distributions respectively get topic-word matrix, topic-topic label matrix, topic-time matrix, calculate topic semantics through each feature matrix; get search item text, text and topic-word matrix to get topic semantics; search item and search item Compute the similarity of the topic semantics; if the similarity meets the set conditions, output the item to be searched. Through the above scheme, a comprehensive and accurate semantic representation can be obtained, and precise semantic search can be realized.

Description

technical field [0001] The invention relates to the technical field of semantic modeling of short texts in social networks, in particular to a semantic search method, system and computer-readable storage medium. Background technique [0002] Nowadays, social networking platforms are developing rapidly, and searching through social networks has become a trend. The text in social network data is very concise. Therefore, the text has semantic sparsity and ambiguity due to different time and location information. Using traditional semantic analysis methods, it is often impossible to obtain accurate semantics. [0003] In order to accurately obtain the semantic representation of short texts in social networks and achieve precise semantic search, the main methods can be divided into two categories. One is to use the co-occurrence frequency of words or the semantic similarity of words to expand short texts into long texts. Words that have nothing to do with short texts will be int...

AI Technical Summary

Problems solved by technology

The text in social network data is very concise. Therefore, the text has semantic sparsity and ambiguity due to different time and location information. Using traditional semantic analysis methods, it is often impossible to obtain accurate semantics.

[0003] In order to accurately obtain the semantic representation of short texts in social networks and achieve precise semantic search, the main methods can be divided into two categories. One is to use the co-occurrence frequency of words or the semantic similarity of words to expand short texts into long texts. Words that have nothing to do with short texts will be introduced in the process of this expansion. When applied to search tasks, the search accuracy cannot be improved.

Another type of method is to build a comprehensive topic model, by constraining the semantic generation process of short texts with semantics in time or geographic location information, so as to obtain short text semantic representations, but semantic sparsity still exists, and, in actual social In network data, geographic location information is difficult to obtain, and at the same time, there is a big problem with the authenticity of geographic location information

[0004] Therefore, when searching for short text semantics by using existing search methods, there are still problems such as short text semantic sparsity and inaccurate search results.

Images