BioGrapher: Biography Questions as a Restricted Domain Question Answering Task
Oren Tsur, Maarten de Rijke and Khalil Sima'an
Proceedings ACL 2004 Workshop on Question Answering in Restricted Domains
This paper addresses Question Answering (QA) for biographical questions, i.e. questions
about biographical facts of persons. The domain of biographical documents
differs from other restricted domains (e.g. computer manuals) in that the available collection
of biographies is rather limited. Hence, a major challenge in biographical QA is to answer
questions about persons whose biographical information is not found in the clean collection.
This paper presents a biographical QA system (called BioGrapher) that addresses this problem
by machine learning algorithms for biography classification. BioGrapher first attempts answering
a question by searching in the collection of biographies (biography.com) using techniques tailored
for the restricted nature of the domain. If a biography is not found, BioGrapher attempts finding
an answer over the internet. Crucially, after document retrieval, BioGrapher filters the retrieved
documents using the biography classifier, and then extracts an answer. We exhibit empirical results
showing that biographical classification, prior to answer extraction, improves the results.