A Comparative Study of Conversion Aided Methods for WordNet Sentence Textual Similarity

摘要

In this paper, we present a comparison of three methods for taxonomic-based sentence semantic relatedness, aided with word parts of speech (PoS) conversion. We use WordNet ontology for determining word level semantic similarity while augmenting WordNet with two other lexicographical databases; namely Categorial Variation Database (CatVar) and Morphosemantic Database in assisting the word category conversion. Using a human annotated benchmark data set, all the three approaches achieved a high positive correlation reaching up to (r = 0.881647) with comparison to human ratings and two other baselines evaluated on the same benchmark data set.