Unknown place name detection base on YamCha for Japanese guidance QA system

We perform a research about Japanese sightseeing guidance question answering (QA) system. In this domain web information retrieval approach is always being used. However for the web information retrieval, Japanese unknown place word detection needs to be considered. Some famous places are known very well, such as "Kinkaku-ji", it can be recognized by morphological analysis. However, the recognition rate of Japanese complex place is so poor. Therefore, we proposed an approach of the Japanese unknown place word detection. We use YamCha, a tool based on Support Vector Machines (SVMs). The goal is that we want to recognize the unknown sightseeing spots word from user's question. In the experiment of Japanese place word detection we have got excellent precision and recalling rates.