Morpheme-based product features categorization in Chinese reviews mining

Pursuing on the analysis of product reviews, an unsupervised product features categorization method is proposed. Morphemes as smallest linguistic meaningful unit are induced in measuring the intra relationship among product features instead of words. Opinion words around product features are chosen to represent the inter relationship among product features instead of full context information. The information gotten from active product features are utilized as constraints to enhance the whole categorization quality. Experimental results show that morpheme has greater linguistic significance than word in Chinese reviews. The opinion words are very important features in measuring the semantic similarity of their associated product features. The proposed method achieves better performance.