Financial named entity recognition based on conditional random fields and information entropy

Shuwei Wang, Ruifeng Xu, Bin Liu, Lin Gui, Yu Zhou

Abstract: Named entity recognition plays an important role in many natural language processing tasks, such as relation detection and information extraction. This paper presents a novel method to recognize named entities in financial news texts in three steps. First, the domain dictionary is applied to recognize stock names. Second, the full form FNEs are identified by incorporating internal features in a classifier based on Conditional Random Fields. Third, the mutual information, boundary entropy and context features are employed to recognize the abbreviation FNE candidates. The experiments completed on a Chinese financial dataset show that the proposed approach achieves 91.02% precision and 92.77% recall.