Text Similarity Calculation Model Based on Semantic Information and Syntactic Structure Fusion Weighting
Zhenhua Zhao, Chao Wang, Shaopei Ji · 2024
Aiming at the problems of traditional text similarity calculation methods that do not fully use the different importance of different parts of the text and ignore the syntactic structure information, this paper proposes a text similarity calculation model based on semantic information and syntactic structure fusion weighting (SSFW). Firstly, a semantic similarity calculation model based on multiple granularity information interactions is designed to accurately extract interaction information and highlight the importance of different granularities. Then, a text similarity calculation model based on Tree-GRU is proposed to perform syntactic structure similarity calculation on the basis of shallow syntactic trees. Finally, the calculation results of semantic similarity and structural similarity are linearly weighted to obtain the text similarity calculation results. The experimental results show that the SSFW model can improve the accuracy and recall rate, which is better than other text similarity calculation methods and more suitable to deal with complex long texts.