A word-based approach for modeling and discovering temporal relations embedded in Chinese sentences

Abstract

Conventional information extraction systems cannot effectively mine temporal information. For example, users' queries on how one event is related to another in time could not be handled effectively. For this reason, it is important to capture and deduce temporal knowledge associated with the relevant events. It is generally acknowledged that information extraction cannot be isolated from natural language processing. As Chinese has no tenses, conventional means for finding temporal references based on verb forms no longer apply. In this article we present an approach for formulating and discovering temporal relations in Chinese. A set of rules is devised to map the combinational effects of the temporal indicators (also known as temporal markers, gathered from various grammatical categories) in a sentence to its corresponding temporal relation. To evaluate the proposed algorithm, experiments were conducted using a set of news reports and the results look promising. Problem discussions are also provided. Through this work, we hope to open up new doors for future research in Chinese temporal information extraction and processing.