Quantifying semantic similarity of Chinese words from HowNet

Semantic similarity is a fundamental concept and widely researched and used in the fields of natural language processing. However, methodologies for measuring semantic similarity are language-dependent. The paper presents a system similarity based measure of semantic similarity for Chinese words from HowNet, an online bilingual (Chinese-English) common sense ontology. The measure is determined in three steps: first, a sememe network is built from concept feature files of HowNet for preparation; then semantic similarity degrees between sememes are given by quantifying their semantic paths in the sememe network, and a sememe weighting method is also provided; finally, a system similarity based semantic similarity degree between Chinese words is presented to combine these elements into a single measure. The experimental results have been adopted by a Chinese query matching system whose precision and flexibility are enhanced thereby.