Polarity Analysis for Food and Disease Relationships

The explosive growth of published articles in biomedical science field has led more research to focus on biomedical relationship extraction. However, there is relatively little investigation conducted on polarity analysis of these relationships, such as food (or nutrition) and disease relationships. In this paper, we investigate how to automatically identify the polarity of relationships between food and disease in biomedical text. In particular, we first analyze the characteristics and challenges of relation polarity analysis, and then propose an integrated approach, which utilizes background knowledge in terms of relation word and polarity class association, and refines this association by using any available domain specific training data. In addition, we propose several novel learning features and a computational approach to construct background knowledge base. Empirical results on real world datasets show that the proposed method is effective.