Building a Knowledge Base for QA System by Linking Korean Vocabulary and Wikipedia

Abstract

For a QA system it's very important to have a notion of vocabulary used in questions and for building correct answers. Especially, when a word represents a concept, one can use related lexical instances to understand it and further extend the knowledge by using the associated information. In this work, we suggest a process of building a knowledge base for such concepts as people, organizations, and places, and linking their instances to Wikipedia articles. We also develop a workbench for KB building. This workbench should efficiently support all features needed to collect necessary data and build the knowledge base. We have created 150,941 links to Korean Wikipedia for 2,394 instances of Korean vocabulary. This KB can be used in QA systems to extend questions, while the workbench can be used to build the KB itself.