Research On Automatic Recognition Of Chinese Time Expression

Temporal information is an important carrier of language semantics, plays a very important role in our lives. People know about event through time, and sort event information in chronological order, grasp the whole process of the development of events. Time recognition as a fundamental task of natural language processing, plays an important role in areas such as machine translation, event detection and tracking, public sentiment topic detection, question answering system, information retrieval and so on.In this paper, a brief introduction and analysis to current research status and available method was brought, methods based on rules, statistics are separately explored to solve the problem of Chinese time expression recognition, and compared the advantages and disadvantages of two methods based on experimental results. On this basis, proposed a generic algorithm for Chinese time expression recognition task based on combining rules with statistics:firstly, analyzed a set of linguistic features of time expression in text such as lexical features and context information, using Conditional Random Fields recognized time unit rather than time expression, avoided the boundary localization problems in Chinese time expressions; obtained the candidate trigger words through the test corpus, scored the candidate trigger words based on evaluation function, filtered out the right time trigger word to perfect time trigger thesaurus; set rules for the time expression boundary localization based on time trigger thesaurus and time affix word thesaurus. Our experimental results show that the F1 value reaches 0.9732 on open test.In this paper, time expression was divided into seven types based on combining the characteristics of the Chinese language and Chinese time expression, namely:DATE, TIME, SET, DURATION, FUZZY, LUNAR, RELATIVE-TIME. Set rules for the features of seven different types after recognizing time expression.