A Study on Traveling Purpose Classification Method to Extract Traveling Requests

Abstract

It is possible to get the flow information of people and transportation efficiently by collecting travel information of people and vehicles in the Internet. In the meantime, Q&A Web sites on the Internet express human requests more directly and they are the useful resources to extract many kinds of knowledge and intentions. This paper proposes the classification method for Japanese speeches in the sites based on the road traffic census by MLIT in Japan to extract traveling requests. Specifically, this method presumes the traveling purposes by SVM. Learning with the frequency of parts of speech that express traveling requests and TFIDF as the features of SVM is carried out. We also evaluated the performance by the experiment and got the result the accuracy 45.5%.