Abstract

During the DARPA Evidence Extraction and Link Discovery (EELD) Program, BBN developed the SERIF system that extracts entities and relations represented in an ontology from text, eg, from web pages, primarily containing news. SERIF's combination of general linguistic models trained on pre-existing corpora with domain specific components trained for the particular task allows powerful linguistic analysis tools to be efficiently applied to relation and event extraction in a new domain. Key to the new strategy was the intermediate use of propositions encapsulating the literal meaning of the text, from which the target relations could then be derived. SERIF has been applied to other evidence extraction tasks, including automated extraction of organizational structure from message traffic and open domain question answering. BBN was one of the groups that achieved the DARPA EELD program goal of 90% of human performance in evidence extraction. Using a hybrid of statistical learning algorithms and handcrafted patterns, SERIF achieved 93% of human performance in extracting entities, events, and relations and 96% of human performance in extracting relations given entities and events. This surpassed our own initial goals, namely to achieve 80% of human performance on entities with names, and 90% on relations when given the entities.

Date

2005

Authors

Elizabeth Boschee, Ralph Weischedel, Alex Zamanian

Journal

Proceedings of the International Conference on Intelligence Analysis

Volume

71