Mining Commercial Networks from Online Financial News

Financial news articles exhibit sufficient information to disclose commercial entities and their behavior. In this paper, a method for commercial network construction is presented, in which natural language processing techniques are applied in commercial entity tagging and commercial relation mining to handle abbreviation recognition, co-reference resolution and contextual commercial relation mining. Preliminary experiments show our method is encouraging. Another contribution of this work is that some fundamental resources are developed for research on commercial network construction. A commercial entity lexicon is manually compiled to collect active commercial entities and a commercial relation lexicon is created to collect keywords that flag commercial relations. Illustration and applications are also discussed, which undoubtedly discloses a promising future of commercial network study.