Relation Extraction from Chinese News Web Documents Based on Weakly Supervised Learning

Extracting instances of a given target relation from a given Web page corpus seems to be the basic work to exploit nearly endless source of knowledge which provided by the World Wide Web. Supervised learning requires a large amount of labeled data, but the data labeling process can be expensive and time consuming. In this paper we present a kernel-based weakly supervised machine learning algorithm for relation extraction. It takes a small set of target relations as input. The goal is to automatically extract arbitrary binary relations from Web documents in the domain of football games. Bootstrapping is used to improve the performance of the system. We also compare the performances on different input example sizes. Experimental results show the effectiveness and benefits of our approach.