Guide word mining method and device, server and storage medium

Abstract

The invention provides a guide word mining method and device, a server and a storage medium, and belongs to the technical field of Internet. The method comprises the following steps: inputting a plurality of to-be-mined corpora into a candidate word mining model, and outputting a plurality of first candidate words and first scores thereof; obtaining a plurality of second candidate words of which the first scores are greater than a first preset score from the plurality of first candidate words; inputting the plurality of second candidate words into the candidate word scoring model, and outputting second scores of the plurality of second candidate words; acquiring a plurality of third candidate words of which the second scores are greater than a second preset score from the plurality of second candidate words; inputting the plurality of third candidate words and the corresponding to-be-mined corpora into a sequence labeling model, and outputting third scores of the plurality of third candidate words; and obtaining third candidate words with third scores greater than a third preset score as a plurality of guide words. According to the method, the mined candidate words are re-scored from the semantic level, the context level and the part-of-speech level, so that the mined guide words are more accurate.

Description

technical field [0001] The present disclosure relates to the technical field of the Internet, and in particular to a guide word mining method, device, server and storage medium. Background technique [0002] With the development of Internet technology and the popularization of mobile smart devices, users are increasingly dependent on the services provided by life service applications for basic necessities of life. Life service applications provide users with massive information on eating, drinking, entertainment, etc. When users use life service applications, life service applications will show users some high-quality guide words, which can help users quickly query Interested information, so as to obtain the required services. [0003] At present, when related technologies mine guide words, the following methods are mainly used: [0004] 1. Mining a plurality of words whose frequency of occurrence is higher than a preset frequency from the corpus to be mined is used as a p...

AI Technical Summary

Problems solved by technology

[0007] However, the candidate positive samples and candidate negative samples used to train the BERT scoring model are phrases with incomplete semantics, which leads to the inaccuracy of the trained BERT scoring model, and simply excavates high-scoring phrases from the corpus to be mined , does not consider the actual context of these phrases, resulting in inaccurate guide words mined

Images