Text similarity computing based on sememe Vector Space

Vector Space Model (VSM) is a classic text presentation model in natural language processing. However the assumption that text terms are pairwise orthogonal is not suitable. General Vector Space Model (GVSM) was proposed to improve the VSM by using term similarity to overcome the pairwise orthogonal term assumption. In this paper, based on GVSM a new approach using HowNet sememe similarity to calculate text similarity in sememe space was proposed and verified by experiment.