Semantic Cosine Similarity

Cosine similarity is a widely implemented metric in information retrieval and related studies. This metric models a text as a vector of terms and the similarity between two texts is derived from cosine value between two texts' term vectors. Cosine similarity however still can't handle the semantic meaning of the text perfectly. This paper proposes an enhancement of cosine similarity measurement by incorporating semantic checking between dimensions of two term vectors. This strategy aims to increase the similarity value between two term vectors which contain semantic relation between their dimensions with different syntax. Experimental result shows our proposal yields a promising result.