Learning for Question Answering and Text Classi cation: Integrating Knowledge-Based and Statistical Techniques

It is a time consuming and diicult task for an individual , a group, or an organization to classify large collections of documents under a content-driven taxonomy. In this paper, we outline an approach for building a system which makes the classiication process the responsibility of the author of the document, thus allowing the author to explain classiications and verify (or correct) automated techniques. We present our preliminary work on such a system, Q&A, which enables the distribution of the task of semantic classiication and knowledge acquisition by semiautomatically learning taxonomic categorizations and document indices as it captures interactions between experts and question-asking users.