Contextual Event Information Extractor for Emails

Abstract

A powerful tool for planning and announcement of Events is Email. Automatic detection of the Occurrence (Title) and its contextual information (Location, Temporal information, Participants) associated with the email is surely desirable to help the users manage and plan important Events. A lot of work has been done in the area of Event detection but it has various limitations from different perspectives. Firstly, the existing work mainly targets text streams like news stories, scientific documents, articles etc that are somewhat structured documents with sufficient event description as compare to the Emails that have structured, semi-structured and unstructured short descriptions with a plenty of description styles. Secondly the objective in most of the research is to detect new or hot events. Thirdly, much of the existing work aims on reporting events and our objective is to support Event Planning and Management. Another lacking thing is the use of publication time as the temporal information instead of actual temporal information contained within text that is indeed required for Event planning and management task. We have used Finite State Automata (FSA) to extract phrases revealing the Places, temporal information and the actual occurrence. The results are evaluated using different measures. Experiments show that the proposed approach performed well on the Email data Corpus.