Speeding Document Annotation with Topic Models

Forough Poursabzi-Sangdeh, Jordan L. Boyd-Graber

Abstract: Document classification and topic models are useful tools for managing and understanding large corpora. Topic models are used to uncover underlying semantic and structure of document collections. Categorizing large collection of documents requires hand-labeled training data, which is time consuming and needs human expertise. We believe engaging user in the process of document labeling helps reduce annotation time and address user needs. We present an interactive tool for document labeling. We use topic models to help users in this procedure. Our preliminary results show that users can more eectively and eciently apply labels to documents using topic model information.