Using Fuzzy Set Similarity in Sentence Similarity Measures

Cross, Valerie, Mokrenko, Valeria, Crockett, Keeley ORCID: https://orcid.org/0000-0003-1941-6201 and Adel, Naeemeh ORCID: https://orcid.org/0000-0003-4449-7410 (2020) Using Fuzzy Set Similarity in Sentence Similarity Measures. In: IEEE World Congress on Computational intelligence - IEEE FUZZ 2020, 19 July 2020 - 24 July 2020, Glasgow, UK (virtual congress).

Preview

Accepted Version Available under License In Copyright. Download (826kB) | Preview

Abstract

Sentence similarity measures the similarity between two blocks of text. A semantic similarity measure between individual pairs of words, each taken from the two blocks of text, has been used in STASIS. Word similarity is measured based on the distance between the words in the WordNet ontology. If the vague words, referred to as fuzzy words, are not found in WordNet, their semantic similarity cannot be used in the sentence similarity measure. FAST and FUSE transform these vague words into fuzzy set representations, type-1 and type-2 respectively, to create ontological structures where the same semantic similarity measure used in WordNet can then be used. This paper investigates eliminating the process of building an ontology with the fuzzy words and instead directly using fuzzy set similarity measures between the fuzzy words in the task of sentence similarity measurement. Their performance is evaluated based on their correlation with human judgments of sentence similarity. In addition, statistical tests showed there is not any significant difference in the sentence similarity values produced using fuzzy set similarity measures between fuzzy sets representing fuzzy words and using FAST semantic similarity within ontologies representing fuzzy words.