CTEMP: A Chinese Temporal Parser for Extracting and Normalizing Temporal Information

R. Dale et al. (Eds.): IJCNLP 2005, LNAI 3651, pp. 694 – 706, 2005. © Springer-Verlag Berlin Heidelberg 2005 CTEMP: A Chinese Temporal Parser for Extracting and Normalizing Temporal Information Wu Mingli, Li Wenjie, Lu Qin, and Li Baoli Department of Computing, The Hong Kong Polytechnic University, Kowloon, Hong Kong {csmlwu, cswjli, csluqin, csblli}@polyu.edu.hk Abstract. Temporal information is useful in many NLP applications, such as information extraction, question answering and summarization. In this paper, we present a temporal parser for extracting and normalizing temporal expres- sions from Chinese texts. An integrated temporal framework is proposed, which includes basic temporal concepts and the classification of temporal expressions. The identification of temporal expressions is fulfilled by powerful chart-parsing based on grammar rules and constraint rules. We evaluated the system on a sub- stantial corpus and obtained promising results. 1 Introduction Temporal information processing is valuable in many NLP applications, such as in- formation extraction, machine translation, question-answering and multi-document summarization. However, a wide scope of linguistic means, from lexical to syntactic phenomena, can represent this information. It is hard to catch the internal temporal meanings which are behind surface texts. The potential applications and the flexibil- ities of temporal representations motivate our research in this direction. In this paper, temporal information is defined as the knowledge about time or dura- tion. This information is crucial for both temporal reasoning and anchoring events on the time line. Temporal expressions are defined as chunks of text which convey direct or indirect temporal information. TIMEX2 annotating guidelines [4, 6] give good descriptions about temporal expressions. According to the guidelines, temporal ex- pressions include dates, times of day, durations, set-denoting expressions, event- anchored expressions, and so on. To retrieve the useful temporal information con- tained in these temporal expressions, we need to identify the extents of temporal ex- pressions in raw text and then represent temporal information according to some stan- dard. The two tasks are called temporal extraction and temporal normalization, re- spectively. We have implemented a full system CTEMP, which consists of two mod- ules: extractor and normalizer. The two modules fulfill temporal extraction and tem- poral normalization, respectively. A comprehensive temporal framework is investigated to analyze the elements in- volved in the mapping procedure, from surface text to internal temporal information. This framework includes basic temporal objects and relations, the measurement of time, and the classification of temporal expressions from Chinese texts. To cope with CTEMP: A Chinese Temporal Parser 695 the flexibilities of the temporal expressions, we have built the temporal parser based on chart-parsing and effective constraints. Experiments with respect to a substantial corpus show that the temporal parser achieves promising results. We took part in TERN 2004 Chinese temporal expression extraction with this temporal parser and our performance is the highest in that track. The rest of the paper is organized as follows: In Section 2 we give a brief discus- sion on related works; Section 3 describes the temporal framework, which is the basis of the whole temporal parser; …

Content truncated to 500 words