Learning event patterns from news text using bootstrapping
by E. Umamaheswari; T.V. Geetha
International Journal of Information and Communication Technology (IJICT), Vol. 7, No. 1, 2015

Abstract: This paper presents a novel method for discovering events described in a news text. In order to determine event specific sentences, it uses a semantic graph representation to express the event semantics. The event specific patterns from the semantic graph are considered as seed examples to identify the example patterns. To extract and learn event patterns, a semi-supervised bootstrapping approach is used for event identification, which requires minimal human supervision. Our aim is to train the event identification system, using a small set of seed patterns for each event, and to find the relationship between the events using instance overlapping between the learned event patterns. Unlike other approaches, this approach uses flexible seed examples, in which it has any number of nodes/concepts and relations. It also differs in acquiring conceptually related event constraints and relations as its features, and additionally, it considers a concept hierarchy of both patterns matching and scoring. We have tested our system on a corpus consisting of 5,000 Tamil news documents, and tested the accuracy of each iteration and achieved an average accuracy of 0.574. We have also tested our approach with the existing MUC-4 dataset and achieved good precision for each event type.

Online publication date: Sun, 30-Nov-2014

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Information and Communication Technology (IJICT):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com