Test Question Classification and Duplicate Checking Based on Course Knowledge Point

Test question duplicate checking by the computer program is the requirement of construction of test question library. To enhance the availability of test question duplicate checking, course knowledge tree and domain term table are introduced in the paper. Based on the domain term table, the domain terms of test question can be easily identified, therefore, the calculating of similarity deals with not only common terms but also domain terms. The traditional TFEDF calculation formula is revised by adding an adjusting factor to ensure that weights of the domain items are not less than an threshold. An improved KNN classifier is proposed that only compute approach degree between the new text and each text of subset of limited training texts that belong to a few of the categories initially identified, so reduces the calculating amount largely.