Improving the Quality of Posts in CQA Websites

Please enable javascript in your browser.

Khatami, Ali | 2020

310 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 53581 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Habibi, Jafar
Abstract:
Every day thousands of questions and answers are posted on CQA websites. Stack Overflow is known as one of the most famous CQA websites in the field of programming. Asking high-quality questions on this website is one of the significant challenges that users face, leading to negative experiences in some cases. In this research history of questions, answers, and post-edits are studied to get insights for predicting questions’ quality. There is a new approach proposed that takes both post scores, and their edits into account to predict their quality. The questions edits that are studied are the one that happened before getting an accepted answer. Furthermore, it is shown that predicting these edits on questions before getting the accepted answer can lead to having a high quality question that recieves its answer much sooner. By predicting the need of edit before posting a question, the user can be sure that she or he is not asking a low-quality question. This prediction is made by a new proposed model which uses NLP and deep neural networks. The model's precision is 91% with 83% recall, which is indeed significant progress concerning recent studies
Keywords:
Data Mining ; Recommender System ; Community Based Question and Answering (CQA) Sites ; Quality Prediction ; Edit Prediction ; Posts Quality

No TOC