Abstract (EN):
In this paper we present the results of applying a statistical query expansion method
on the retrieval stage of a QA system for Portuguese (RAPOSA). Our approach in-
volves expanding queries for event-related or action-related factoid questions using a
verb thesaurus automatically generated using information extracted from large cor-
pora. We show that our expansion approach improves QA recall when compared with
applying expansion based on a simple form of stemming, while simultaneously requir-
ing the analysis of only 30% as many text snippets. However, we were not able to
outperform the recall obtained using an even simpler expansion method, which never-
theless achieves lower precision and requires analyzing many more text snippets. We
conclude by observing that a more thorough analysis of the usefulness of our approach
on QA performance requires improving other stages of the QA pipeline which currently
impose significant limitations on the overall performance of the system.
Language:
English
Type (Professor's evaluation):
Scientific