Extracting Event Temporal Information Based on Web

Abstract

Temporal information is an important characteristic of event. It can be used in information retrieval process to organize the returned result. In Chinese, the presentations of time expression are very complex, which make it difficult to both accurately recognize a time expression and precisely connecting it with a given event in a web page that contains multiple events. To address these problems, this paper presents an innovative event time extraction model. Rather than just rely on local context within a web page or a text, this model applies global context provided by all the web pages that had been automatically judged as related with a given event. Our experiment results based on the evaluation criterions show the feasibility of the provided model.

Boundary detection of multiple related temporal duration of schedules in email

“…[8] tried to apply conditional random fields to this problem. Also, there were some researches to extract time expressions from complex contexts like emails or web pages, for example [10] or [4], but they rarely move onto the classification of time expressions into starting time and ending time. [3] classified time expressions into stime and etime by first doing lightweight parsing on the sentence and then matching text terms to each field.…”