IExM: Information Extraction System for Movies

Peng-Yu Chen, Yi-Hui Lee, Yueh-Han Wu, Wei-Yun Ma

Abstract: In this demonstration, we present Information Extraction System for Movies(IExM), which helps extract relation instances from un- labeled movie articles. We have designed a new distant-supervised learning algorithm: Improved Pattern Ranking Algorithm(IPRA) to extract relation instances from unlabeled articles, which iteratively generates new patterns starting from a limited set of seed instances, and extracts new instances using high-ranking pattern in a precise and effective way. IPRA also has a special estimation for the newly generated patterns based on the quality estimation of the instances that generate the patterns and ranks patterns' quality based on var- ious factors.