A novel locality-sensitive hashing relational graph matching network for semantic textual similarity measurement. (30th November 2022)

• Record Type: Journal Article
• Authors: Li, Haozhe; Wang, Wenhai; Liu, Zhaoran; Niu, Yunlong; Wang, Hao; Zhao, Shunping; Liao, Yilin; Yang, Weigeng; Liu, Xinggao
• Is Part Of: Expert systems with applications. Volume 207(2022)
• Publication Date: 2022-11-30
• Subjects: Semantic textual similarity -- Graph neural network -- Natural language processing -- Deep learning; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2022.117832
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: Semantic textual similarity measurement depends on the syntax. Pruned syntactic dependency graph performs better than original tree. The most expensive part of inference time is the interactions of words. Locality-sensitive hashing mechanism is introduced into interactions of words. LSHRGMN both encodes the syntactic dependency graph and interactions of words. Abstract: Recent efforts adopt interaction-based models to construct the interaction of words between sentences, which aim to predict whether two sentences are semantically equivalent or not in semantic textual similarity (STS) task. However, these methods lack the global semantic awareness, which make it difficult to distinguish syntactic differences and also suffer from the inference time cost, primarily due to the calculation of the pair-interactions of words. A novel model called Locality-Sensitive Hashing Relational Graph Matching Network (LSHRGMN) is therefore proposed, which tackles these problems by syntactic dependency graph and locality-sensitive hashing (LSH). Specifically, syntactic dependency graph is aware of the global semantic information via rooting in each word to construct several trees and merging all the trees into one graph. LSH mechanism is introduced into pair-interactions of words for the inference efficiency problem. Extensive experiments are conducted on three real-world datasets, and the result shows that the proposed approach acquires higher accuracy and intriguing inference speed.