Knowledge-based sentence semantic similarity: algebraical properties

Abstract Determining the extent to which two text snippets are semantically equivalent is a well-researched topic in areas of natural language processing, information retrieval and summarization. The sentence-to-sentence similarity scoring extensively used both generic query-based summarization documents as significance or indicator. Nevertheless, most these applications utilize concept semantic measure only tool, without paying importance inherent properties such tools that ultimately restrict scope technical soundness underlined applications. This paper aims contribute fill this gap. It investigates three popular WordNet hierarchical measures, namely path-length, Wu Palmer Leacock Chodorow, from algebraical intuitive properties, highlighting their limitations theoretical constraints. We have especially examined related range score, incremental monotonicity evolution, with respect hyponymy/hypernymy relationship well set interactive properties. Extension word sentence has also been investigated using pairwise canonical extension. Properties scrutinized. Next, overcome terms accounting for various Part-of-Speech categories, “All word-To-Noun conversion” makes use Categorial Variation Database (CatVar) put forward evaluated publicly available dataset comparison some state-of-the-art methods. finding demonstrates feasibility proposal opens up new opportunities processing tasks.