Enhancing Short Text Semantic Similarity with Word2Vec and Maximum Matching

In natural language processing, Short Text Semantic Similarity (STSS) is a highly prominent field. It has significant impacts on a wide range of applications, such as sentence classification, natural language inference, paraphrase identification, question answering, sentiment analysis, and semantic relevance scoring, and more. In this paper, we propose a short text similarity calculation model that combines Word2Vec and the maximum matching algorithm. Our experiments on short text semantic relatedness scoring using the publicly available SICK (Sentences Involving Compositional Knowledge) dataset showed encouraging results. Compared to other models, the model proposed in this paper achieved good performance, with higher overall Pearson and Spearman correlation measures between predicted similarity scores and human ratings than other models.