Premium
Identifying the reasons contributing to question deletion in educational Q&A
Author(s) -
Rath Manasa,
Shah Chirag,
Floegel Diana
Publication year - 2017
Publication title -
proceedings of the association for information science and technology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.193
H-Index - 14
ISSN - 2373-9231
DOI - 10.1002/pra2.2017.14505401036
Subject(s) - order (exchange) , computer science , meaning (existential) , quality (philosophy) , typology , construct (python library) , process (computing) , knowledge base , world wide web , psychology , data science , internet privacy , sociology , epistemology , philosophy , finance , anthropology , psychotherapist , programming language , operating system , economics
Community question‐answering (CQA) services are widely used by information seekers looking to ask questions and obtain accurate, personalized answers. Though general CQA sites such as Yahoo! Answers attract a diverse pool of users from many walks of life, other sites cater to a specific user pool. While identifying bad CQA content is generally important in order to improve sites' overall health and community knowledge‐sharing, examining educational CQAs is particularly urgent in order to help struggling students understand why their questions fail, re‐frame their inquiries in a more accurate manner based on feedback, and ultimately receive correct answers that facilitate their learning process. Otherwise, students' questions would merely be deleted, meaning they lose multiple opportunities to enrich their knowledge base. In this work, we focus on questions posted to Brainly, the largest educational CQA site, in order to first identify “bad” questions and next understand what textual (content‐based) features contribute to such questions' poor quality. Using a sample of 1,000 questions–500 of which were deemed “good” and 500 of which were deemed “bad” by site moderators– we attempt to automatically classify question quality in order to label which questions would be deleted and therefore go unanswered. We then use human assessment to expand upon a typology to classify poor quality questions based on 14 textual features in order to identify why they have been marked for deletion. Finally, we propose a method to automatically identify questions' problematic textual features in order to provide feedback to students posting “bad” questions and ensure that they are given the opportunity to revise and improve their inquiries to obtain accurate answers that resolve their information needs.