| Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 94 === Question Answering System is expected to response exact answers to users’ questions in natural languages. Typically, most existing QA systems consist of the following components: question analysis, information retrieval, and answer extraction. This paper focus on question classification of question analysis, and we proposed a new method to deal with this problem. Recently, some machine learning techniques like support vector machines are employed for question classification. However, these techniques heavily depend on the availability of large amounts of training data, therefore if there is not enough training data, it may make an ineffective result. We think that in addition to question focus, there are some useful dependency features in a question, and these dependency features can be helpful for question classification.
In this paper, we present a simple learning method that explores Web search results to collect more training data automatically, and we also proposed two models, the first is Dependency Feature Model (DFM) which takes advantage of dependency features learned from the larger number of collected Web search results to support the determination of question type, the second is Dependency Relation Model (DRM) which used dependency relations between question focus and dependency features to support the determination of question type.
|
|---|
