Abstract

News articles are a particular type of story that reflects our daily life on various media. The most distinctive characteristics of news story are its discontinuity in the ordering of events and its limited shelf life. These characteristics present difficulties for automated systems which deal with temporal information. Such applications include information extraction, question answering, summarization, machine translation, etc. In recent years, some concrete work has been done in the area of event identification and temporal expression annotation. Some issues still remain regarding implicit temporal information. This major report describes a method to find patterns in news stories that journalists use to organize events. The study will help us better understand the nature of news stories. It may also help to get implicit temporal information by the implication from the event in the same paragraph. The major report also presents an annotation scheme for annotating temporal expressions and paragraphs in new texts. In this study, we manually annotated all temporal expressions and paragraphs in selected news texts. We build the time structure for each of the texts according to the story time of the paragraphs. After all the target news texts are analyzed and annotated, we compare time structures against each other to derive time structure patterns that may exist in news stories.