An unsupervised center sentence-based clustering approach for rule-based question answering

摘要

Question answering (QA) systems have widely employed clustering methods to improve efficiency. However, QA systems with unsupervised automatic statistical processing do not seem to achieve higher accuracies than other approaches. Therefore, with the motivation of obtaining optimal accuracy of retrieved answers under unsupervised automatic processing of sentences, we introduce a syntactic sequence clustering method for answer matching in rule-based QA. Our clustering method called CEnter SEntence-baseD (CESED) Clustering is able to achieve accuracies as high as 84.62% for WHERE-type questions.