Ontological and Fuzzy Set Similarity between Perception-Based Words

Fuzzy short text semantic similarity measures allow the inclusion of human perception based words to be within the similarity measurement which results in better correlation on the meaning of the short text with human understanding. Existing measures such as FUSE and FAST rely on the creation of fuzzy ontological structures from the modelling of perception words using type-1 or type-2 fuzzy sets. Due to the complex methodology of creating these ontologies, fuzzy word representation cannot be guaranteed due to language evolution. This paper presents a comparative study of simpler fuzzy set similarity measures. The results surprisingly indicate that a very simple fuzzy set similarity measure created from the center of gravity (COG) distance between type-2 fuzzy sets has a very high correlation with the FUSE semantic similarity measure.