TEXT MINING ON Q&A WEBSITES

Abstract

A well-known online Q&A forum where software developers post and answer questions related to programming. We perform the analyses on a long list of factors in the raw data and identify those that have a clear relation to response time. The main search is for the tag-related factors, such as their count (frequency of the tag used) and the cardinality of their “subscribers” (number of users can answer questions containing the tag), provide much stronger confirmation that factors not related to tags. Finally, we learn models using the identified evidential features for predicting the response time of questions, which also manifest the significance of tags chosen by the questioner. The process of defining the factors incorporates different steps for the analysis of documents like a social post or raw data (extraction of keywords and their matching to the related concepts) and their weighting. The concept of weighting involves different scores, such as statistical patterns and sentiment analysis which attempt to measure the proficiency of the user in the relative field.