ECAI 2004 Conference Paper

[full paper]

Answer Type Checking in Open-Domain Question Answering

Stefan Schlobach, Marius Olsthoorn, Maarten de Rijke

Open domain question answering systems have to bridge the potential vocabulary mismatch between a question and its candidate answers. One can view this as a recall problem and address it accordingly. Recall oriented strategies to question answering may generate considerable amounts of noise. To combat this, many open domain question answering systems contain an explicit filtering or re-ranking component. In many cases this involves checking whether the answer is of the correct semantic type. We compare and explore redundancy-based and knowledge intensive strategies for this "answer type checking". Common to both approaches is that the semantic types are structured as concepts in an ontology of the domain for a particular question type. We implement these strategies for geographical questions and evaluated on over 900 questions. Overall, we find that type checking, independent of the chosen strategy, improves the number of correct answers significantly. Moreover, each of the described approaches improve on questions where the other fails. As knowledge intensive approaches are labor intensive in the creation and domain dependent, but efficient, whereas redundancy-based methods are more easily extensible to other domains, but sometimes inefficient, both methods have their merits.

Keywords: Information Extraction, Ontologies, Information Retrieval, Text Mining

Citation: Stefan Schlobach, Marius Olsthoorn, Maarten de Rijke: Answer Type Checking in Open-Domain Question Answering. In R.López de Mántaras and L.Saitta (eds.): ECAI2004, Proceedings of the 16th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2004, pp.398-402.