Question Answering System Using Concept-Based Vector Space Model

This paper presents the architecture of the Concept-based Vector Space Model Question Answering System (CBVSM-QAS) developed at the Nagaoka University of Technology (NUT) and used in the 4-th NTCIR workshop Question Answering Challenge (QAC) evaluation. The CBVSM-QAS runs on the factual question, which is, correspond to the subtask-1 in the NTCIR-4’s QAC. One major peculiarity of this system is the idea of placing the whole data set in the concept-based vector space, and the searching of the answer for each question is done by the calculating the nearest newspapers documents’ vector. In this paper, the architecture of the system used during the formal run of NTCIR-4’s QAC and the architecture of the system after some improvement is introduced.