Exploiting Syntactic and Semantic Information for Textual Similarity Estimation

Abstract

The textual similarity task, which measures the similarity between two text pieces, has recently received much attention in the natural language processing (NLP) domain. However, due to the vagueness and diversity of language expression, only considering semantic or syntactic features, respectively, may cause the loss of critical textual knowledge. This paper proposes a new type of structure tree for sentence representation, which exploits both syntactic (structural) and semantic information known as the weight vector dependency tree (WVD-tree). WVD-tree comprises structure trees with syntactic information along with word vectors representing semantic information of the sentences. Further, Gaussian attention weight is proposed for better capturing important semantic features of sentences. Meanwhile, we design an enhanced tree kernel to calculate the common parts between two structures for similarity judgment. Finally, WVD-tree is tested on widely used semantic textual similarity tasks. The experimental results prove that WVD-tree can effectively improve the accuracy of sentence similarity judgments.

A Cross-Lingual Sentence Similarity Calculation Method With Multifeature Fusion

“…They argued that the traditional vector space takes a single word as the linguistic unit of a sentence and only considers the statistical information of the word without considering the semantic information of the word, so their proposed improved algorithm takes the concept as the minimum unit of the sentence, improves the accuracy of the sentence similarity algorithm through concept abstraction and specialized classification, and achieves good results in question and answer systems. Luo et al [34] proposed a method for text similarity computation using syntactic and semantic information that considers weight vector dependency trees as both the semantic information and structural information of sentences. Ruan et al [35] considered sentence morphological information but did not consider other semantic-related information of sentences.…”