Chinese News Event 5W1H Elements Extraction Using Semantic Role Labeling

To relieve "News Information Overload", classification, summarization and recommendation techniques have been proposed. However, these techniques fail to provide sufficient semantic information about news events. In this paper, considering5W1H (Who, What, Whom, When, Where and How), the full list of elements of a news article, we propose a novel approach to extract event semantic elements. The approach comprises a key event identification step and an event element extraction step. We first use machine learning method to identify the key events of Chinese news stories. Then we employ semantic role labeling (SRL) enhanced by heuristic rules to extract event 5W1Helements. A prototype system is implemented based on proposed approach. Extensive experiments on real online news data sets confirm the reasonability and feasibility of our approach.