Understanding Question Quality through Affective Aspect in Q&A Site
Jirayus Jiarpakdee (1551124)
Internet allows people to easily exchange and share knowledge.
Through the World Wide Web, one of the most widely used knowledge sharing service is a question and answering site.
However, not every question is answered, up to 42.47% in StackOverflow.
To solve unanswered question problem, several approaches have been proposed.
StackOverflow web developers and Eric S. Raymond suggested how users should improve their questions.
In addition, researchers published approaches to study the question quality using contents of a question and answers, community-based features that represent relations among users in the community, and affective features that reveal emotional information.
Although emotional information is widely used and included in natural language processing field, little has been studied in question and answering mining research domain.
In this thesis, we investigated (1) the relation between affective features and the prediction of question quality, and (2) the impact that affective features have on question quality.
Through the analysis on MSRChallenge2015 dataset, we found that using affective features improves the performance of question quality prediction.
Moreover, while Favorite Vote Count feature has the highest impact on our prediction models, Politeness, one of the affective features, is ranked in the second highest group of features.