A Topic Based Method to Classify the Question Clarity in CQA Networks

Abstract

Better results would be obtained by distinguishing the clarity of questions as well as increasing their quality. This improvement in clarity may improve the output of a search engine when it encounters the query. Furthermore, it can lead to getting the correct answer to a question when asked in CQAs. In this regard, thousands of different questions are posted daily in CQAs, making these questions and their answers one of the world's most valuable information sources. Nonetheless, most of the questions posted in these forums do not result in proper answers, with one of the most important reasons being a lack of clarity in the questions. This paper addresses one of the most important issues in this field, which is classifying questions based on their clarity. For this purpose, a feature vector based on clustering approaches and obtaining similar questions is designed uniquely for each question based on the total data provided in this field. Following that, the questions are classified based on their clarity using a machine learning classification model. Furthermore, we investigated and reported our new approach using other related approaches in this field in the following step. What we describe as an accomplishment in this paper is the high separability of the questions based on the feature vector extracted by different clusters, which has a much higher performance when compared to other proposed textual classification methods.