A Semantic Level Matching Method for Unstructured Documents
Download as PDF
DOI: 10.25236/iccem.2021.044
Author(s)
Liu Hai, Chen Xiaoming
Corresponding Author
Liu Hai
Abstract
This paper proposes a text semantic matching model, designs multiple features for text matching from word level, phrase level, sentence level and semantic level, and uses ranking learning method to fuse features. Experiments are carried out to verify the performance of the proposed method in text matching and real scene human-computer dialogue tasks. Experiments show that the multi granularity text semantic matching model can achieve high accuracy in text matching task, and has good domain migration ability.
Keywords
Semantic matching, Unstructured, Multi granularity