Language model-based sentence classification for opinion question answering systems

摘要

In this paper, we discuss an essential component for classifying opinionative and factual sentences in an opinion question answering system. We propose a language model-based approach with a Bayes classifier. This classification model is used to filter sentence retrieval outputs in order to answer opinionative questions. We used Subjectivity dataset for our experiments and applied different state-of-the-art smoothing methods. The results show that our proposed technique significantly outperforms current standard classification methods including support vector machines. The accuracy is improved from 90.49% to 93.35%.