Mining Unstructured Knowledge Bases for Question & Answers Websites

Knowledge retrieval is a long standing problem that faces the challenge of dealing with natural language data. Such data is produced and consumed by humans and therefore is ambiguous and unstructured by nature. Recent studies focused on using unstructured data as sources for answers with interesting results. We used Stack Overflow public data and searched for answers in external sources. Results showed that this approach is feasible. Furthermore we explored machine-learned models and were able to improve the baseline results by leveraging the tag information predicted from new questions.