Research On Temporal Relation Between Time Expressions And Events In Chinese Language

With the rapid development of information technique, the unstructured source onInternet has an exponential growth. The demand of effectively dealing with these datapromoted the emergence of information extraction. Information Extraction (IE) is thename given to any process which selectively structures and combines data which isfound, explicitly stated or implied, in one or more texts. The automatic identification ofall temporal relations between time expressions and events is the ultimate aim ofresearch in this area. The task is to determine the relation between an event and a timein the same sentence. This capability is crucial to a wide range of NLP applications,from document summarization and question answering to machine translation etc.A better time expressions recognition and events extracting is the foundation ofresearch on temporal relation between time expressions and events. However, theexisting research prefers focusing on words’ structure information which is the basicfeature extracted in time expression recognition and event extracting. In order to addressthis issue, this paper has proposed a few novel approaches, which have taken syntacticinformation and semantic information into account. Meanwhile, these approaches havebeen applied in the automatic identification of all temporal relations between timeexpressions and events. All these research works can be concluded as follows:1. The approach for automatic identification of Chinese time expressionFirstly, we annotated the time expressions, and then classified the annotated timeexpressions; ultimately the first two results composed the time expressions. In theprocessing of annotating and classifying the time expressions, words’ information iswidely used. Consider the semantic level character of time expressions whoseannotating and classifying cannot rely on the words level features. In order to addressthis issue, this paper proposed an approach which has taken features which extractedfrom syntactic information and semantic information into account and an effectivealgorithm was proposed to extract the corresponding features. Moreover, an approach ofautomatic identification of Chinese time expression was proposed based on the effectivefeature. Experiment results demonstrated that our approach have received a betterperformance.2. The approach for automatic identification of Chinese eventsEvents extraction means extracting the words and phrases which represent the situations that happen or occur, or describe states or circumstances in which somethingobtains or holds the truth in a sentence. Generally, the researchers just extract the words’information in event recognition since they only treat a single word as an event.Considering not only single word but also phrase can represent an event, this paperextract the syntactic information in Chinese events extraction. Experiment resultsdemonstrated the improvement of accuracy. As the semantic information represents ashallow grammatical structure of a sentence, which not only implies semanticrelationships between the predicate and other words in the predicate framework, alsoimplies semantic relationships of different words in the predicate framework, Therefore,we exact the semantic information in events extraction. This paper proposes anapproach of Chinese events extraction based on features extract from words’information、 syntactic information and semantic information. Experiment resultsdemonstrated that our approach have received a better performance.3. The approach for automatic identification of all temporal relations between timeexpressions and eventsBased on the former two researches, this paper extended to propose an approach ofautomatic identification of all temporal relations between time expressions and events.Experiment results demonstrated that our approach have received a better performance.