Classifying Questions Into Fine-Grained Categories Using Topic Enriching

The lasting popularity of many social Q&A websites, such as Yahoo! Answers and ResearchGate, has become valuable knowledge repositories for people to search for answers to questions in various aspects in life. Finding the most relevant questions is often a non-trivial task, and a fine-grained classification system of questions will be an important aid. Existing work mainly focused on classifying questions into different major categories (e.g., "Health","Computer", etc.) without further dealing with the fine-grained categories (e.g., "Dental","Skin Conditions", etc.). Identifying questions' finegrained categories is challenging due to the limited length of a question and insufficient content information available in these social Q&A websites. In this work, we propose a novel framework to classify questions into fine-grained categories based on enriching the related topics of questions. We leverage word embedding feature representation with topic modelings to determine the extended feature terms, i.e., terms that do not appeared in original question content. The enriched features then are used for fine-grained category classification. Extensive experiment results based on three large data collections showcase the effectiveness of our proposed approach.