Classifying stack overflow questions based on bloom’s taxonomy

Abstract

Bloom’s Taxonomy is a framework which acts as a reference for classification of questions across different cognitive levels such as Knowledge, Comprehension, Application, Analysis, Synthesis, and Evaluation. It can be used to select questions in order to evaluate knowledge and understanding of students. We, in this thesis, work on the problem of knowledge management and try to classify questions asked on popular social networks like Stack Overflow (SO). The motivation for the problem comes from the SO being as one huge source of technical questions and answers which include current trending discussions also. Such a knowledge source can be very useful for the education domain. We first apply LDA to reduce the dimensions of each SO document and then use k-means algorithm on a collection having unlabeled and labelled documents to get the result. We obtain an accuracy of 30.2% with this approach. We further augment other features like score, answer count and view count to the obtained feature set and get an accuracy of 56.33%.