engleski

Question Classification for a Croatian QA System

Question Answering (QA) systems provide efficient means for retrieval of information, which in many cases more directly address users' information needs. The performance of a QA system crucially depends on its ability to correctly classify the query question according to the expected answer type. This paper addresses the problem of a question classification for the Croatian language, as a first step towards building an open-domain QA system. We compare different machine learning classifiers on a Croatian test collection based on a two-level question taxonomy. The evaluation results are encouraging and comparable to state- of-the-art results for other languages: accuracy is over 80% for coarse-grained classification and almost 70% for fine-grained classification.

Question classification; question answering; information retrieval; Croatian language

nije evidentirano