Similarity search method of enterprise names

Abstract

The invention relates to a similarity search method of enterprise names. The similarity search method comprises steps as follows: input search keywords are decomposed, and the processed search keywords are obtained, wherein the search keywords are to-be-researched enterprise names; search phrases are determined according to the processed search keywords; similarity search is performed on the determined search phrases and search results are obtained; the enterprise names ranking in the top N of the search results are displayed to be checked by users, and N is an integer larger than 1. With the adoption of the method, the search efficiency is greatly improved, the users can check similarity research results, and the requirement of similarity search business is met.

Description

technical field [0001] The invention relates to the technical field of similarity retrieval, in particular to a method for similarity retrieval of enterprise names. Background technique [0002] The precise query speed of the database is quite fast, but the fuzzy query speed will decrease rapidly after the amount of data exceeds one million, especially for the "contains" relationship, usually exceeding 10 seconds. For example: if you want to hit "Great Wall Computer Software" by entering the keyword "computer", the speed will be very slow. What's more, according to the rules for judging the similarity of names, it is necessary to use circular nested fuzzy queries, and the performance is completely unacceptable. [0003] Usually when encountering this kind of problem, "full-text search technology" will be used, such as: Baidu, Sogou, etc., to achieve fast search results in massive data based on a small number of keywords. However, the inventor found in the process of implem...

AI Technical Summary

Problems solved by technology

[0005] Secondly, general-purpose full-text search engines can also support word-by-word search, but the performance advantage is lost. For example, for a 15-character name string, the conclusion obtained by searching according to the full arrangement of words is: only by " "Full-text search" can't fully realize the business requirements for approximate retrieval of enterprise names. The search time limit will exceed 30 seconds, and the sorting of the retrieval results is not accurate, which is quite different from the approximate name that people usually feel, which is far from meeting the business requirements.

Images