Improved Sentence Similarity Measurement in the Medical Field Based on Syntactico-Semantic Knowledge

Abstract

AbstractComputing semantic sentence similarity plays a vital role in a range of text mining applications. In the clinical domain, Semantic Textual Similarity can enable us to detect and eliminate redundant information that may lead to a reduction in cognitive burden and an improvement in the clinical decision-making process. Several methods have been proposed to measure the sentence Similarity based on semantic knowledge and learning models. Despite realized efforts, the results of these methods are unsatisfactory, as much relevant semantic knowledge, such as semantic class, thematic role and syntactico-semantic knowledge like the semantic predicates, are not taken into account. In this paper, we propose a novel method to measure semantic similarity between clinical sentences based on deep learning and using syntactico-semantic knowledge such as semantic argument and thematic role. An experiment was carried out on MedSTS dataset yielded better results, showing a high correlation (r = 0 89) with human ratings.KeyswordsSemantic similarityClinical sentenceSyntactico-semantic knowledgeDeep learning