Credit Distribution: A Graph-Based Approach to Extract Product Description Words
Songxiang Cen, Yu Liang Mao, Rongjun Li, Xiaojie Wang · 2008
In this paper, we proposed a method called "Credit Distribution" to extract product entities and features from Chinese customer reviews. Given several seed words, the system can discover new entity words and feature words appeared in the corpus. Quite different from the previous work, the proposed method first draws the graph of words in the corpus and then distributes the credit from the given seed words based on the graphpsilas topology structure. Finally, high ranked words are extracted. Our experiment results show that the method proposed here is highly effective.