Applying Machine Learning to Chinese Temporal Relation Resolution

Applying Machine Learning to Chinese Temporal Relation Resolution Wenjie Li Department of Computing The Hong Kong Polytechnic University, Hong Kong cswjli@comp.polyu.edu.hk Kam-Fai Wong Department of Systems Engineering and Engineering Management The Chinese University of Hong Kong, Hong Kong kfwong@se.cuhk.edu.hk Guihong Cao Department of Computing The Hong Kong Polytechnic University, Hong Kong csghcao@comp.polyu.edu.hk Chunfa Yuan Department of Computer Science and Technology Tsinghua University, Beijing, China. cfyuan@tsinghua.edu.cn Abstract Temporal relation resolution involves extraction of temporal information explicitly or implicitly embedded in a language. This information is of- ten inferred from a variety of interactive gram- matical and lexical cues, especially in Chinese. For this purpose, inter-clause relations (tempo- ral or otherwise) in a multiple-clause sentence play an important role. In this paper, a computa- tional model based on machine learning and heterogeneous collaborative bootstrapping is proposed for analyzing temporal relations in a Chinese multiple-clause sentence. The model makes use of the fact that events are represented in different temporal structures. It takes into ac- count the effects of linguistic features such as tense/aspect, temporal connectives, and dis- course structures. A set of experiments has been conducted to investigate how linguistic features could affect temporal relation resolution. 1 Introduction In language studies, temporal information de- scribes changes and time of changes expressed in a language. Such information is critical in many typi- cal natural language processing (NLP) applications, e.g. language generation and machine translation, etc. Modeling temporal aspects of an event in a written text is more complex than capturing time in a physi- cal time-stamped system. Event time may be speci- fied explicitly in a sentence, e.g. “他们在 1997 年解 决了该市的交通问题 (They solved the traffic prob- lem of the city in 1997)”; or it may be left implicit, to be recovered by readers from context. For example, one may know that “修成立交桥以后，他们解决了该 市的交通问题 (after the street bridge had been built, they solved the traffic problem of the city)”, yet without knowing the exact time when the street bridge was built. As reported by Partee (Partee, 1984), the expression of relative temporal relations in which precise times are not stated is common in natural language. The objective of relative temporal relation resolution is to determine the type of rela- tive relation embedded in a sentence. In English, temporal expressions have been widely studied. Lascarides and Asher (Lascarides, Asher and Oberlander, 1992) suggested that tempo- ral relations between two events followed from dis- course structures. They investigated various contextual effects on five discourse relations (namely narration, elaboration, explanation, back- ground and result) and then corresponded each of them to a kind of temporal relations. Hitzeman et al. (Hitzeman, Moens and Grover, 1995) described a method for analyzing temporal structure of a dis- course by taking into account the effects of tense, aspect, temporal adverbials and rhetorical relations (e.g. causation and elaboration) on temporal order- ing. They argued that rhetorical relations could be further constrained by event temporal classification. Later, Dorr and Gaasterland (Dorr and Gaasterland, 2002) developed a constraint-based approach to generate sentences, which …

Content truncated to 500 words