A Method for Recognition of User Search String Organization Name Based on Semantic Feature Model

Abstract

The invention belongs to the field of natural language processing, and in particular relates to a method for identifying user search string institution names based on a semantic feature model. It includes the processing of the model building phase and the recognition phase. In the model building stage, the existing long text annotation corpus is used to build a training corpus that conforms to the distribution of user search strings. This corpus not only preserves the traditional features of word segmentation and part-of-speech tagging, but also adds contextual features and stickiness in query strings. The combined degree feature is related to the semantic environment feature, and the conditional random field model is established according to the compound semantic feature as the organization name recognition model. In the identification stage, the pattern sequence of the user query string is obtained by calculating the corresponding semantic environment features of the user search string, and the pattern sequence matching the organization name is extracted, that is, the organization name in the user search string is obtained. The method comprehensively improves the accuracy rate and recall rate of identifying the organization name in the user search string.

Description

technical field [0001] The invention belongs to the field of natural language processing, and in particular relates to a method for identifying user search string institution names based on a semantic feature model. Background technique [0002] Today's society has become an era of information explosion, and the rapid development of the Internet has enabled China to have more than 600 million Internet users and ZB (ZettaByte)-level data accumulation. The search engine greatly facilitates people's daily life, work, and learning information acquisition needs, and the importance of information screening and sorting is particularly prominent. After the user enters their own questions into the search engine, the search engine will perform a series of preprocessing links such as word segmentation, stop word removal, error correction, and entity recognition on the user search string. In these preprocessing links, each item Both are extremely important and indispensable. The qualit...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to solve the problem of low accuracy when the existing long text organization name recognizer is used for the organization name recognition of the user search string because the user search string is relatively short and lacks a semantic environment. Recognition Method of User Search String and Organization Name Based on Feature Model

Images