Neural Networks for Explanatory Sentence Recognition

Opinion mining has been received extensive interests in natural language processing community. Previous work drives much attention on the polarities of opinions, as well as their holders and targets. Little work concerns the reason that lead opinion holders express such opinions, which are also as important for opinion mining. In this work, the first neural models for explanatory sentence recognition is proposed, detecting whether a given opinionated sentence contains the explanatory information for the expressed opinion. Experimented results show that the proposed neural models achieve better performances than baseline discrete models on two Chinese datasets.