Word-level neutrosophic sentiment similarity

In the specializedliterature, there are many approaches developed for capturing textual measures: textual similarity, textual readability and textual sentiment. This paper proposes a new sentiment similarity measures between pairs of words using a fuzzy-based approach in which words are considered single-valued neutrosophic sets. We build our study with the aid of the lexical resource SentiWordNet 3.0 as our intended scope is to design a new word-level similarity measure calculated by means of the sentiment scores of the involved words. Our study pays attention to the polysemous words because these words are a real challenge for any application that processes natural language data. After our knowledge, this approach is quite new in the literature and the obtained results give us hope for further investigations.