Abstract: In Chinese-English machine translation, temporal information parsing of Chinese text is very important, because it is the basis to generate the correct tenses of English verbs. Note that this tense is a syntactic category of English, and there is another tense in this paper. The second tense is one of the most important elements of temporal information. It is a temporal relation between the event time and the speech time or another reference time, and can be obtained by time phrase parsing of Chinese text. In this paper, time phrases of Chinese text are classified firstly, then a time phrase semantic representation structure （TPSRS） is designed to represent semantic of time phrases, and concept information unit relation network （CIURN） is used to represent context knowledge. Therefore an algorithm TPPA is proposed to parse time phrases. With semantic of time phrases, tense calculus is defined to infer tenses of verbs of Chinese texts. Finally, this approach is implemented in the Chinese-English machine translation system ICENT, and the results of the experiment are satisfying for those Chinese texts that have the written time.