Mining consumer´s opinion target based on translation model and word representation

In recent years, the E-commercial plays an important role in people´s daily life. When on the Internet, people often buy commodities from Taobao, Tmall and make comments on them, the comments of the goods may have closely connections with commercial value, which often reflect what´s the consumers really care when they choose one piece of good among thousands of other similar ones. How to mining these aspects which the consumers really concern is a problem left unsolved. As a potential effective solution to construct structured information for people´s preference, Information Extraction(IE) has attracted more and more scholar´s attention. A meaningful research area is Opinion Target Extraction(OTE). This paper proposed a system using translation model as well as word representation method to obtain user´s interests on dataset in Chinese. To release the word segmentation error, a finely generated system with new Chinese word detection module is proposed. The experiments on two corpus subjected on digital product verify the effective of our method.