부산대학교 도서관

We propose a new model for estimating the place of residence of Twitter users living in Japan. Some Japanese users tend not to post geotagged tweets and tweets containing characteristic words related to their residential areas, and not to describe the accurate location in their user profiles. Therefore, it is necessary to propose a method to estimate the user's place of residence using the information other than geotagged tweets and user profiles together. In this paper, we propose a deep neural network model that estimates user's place of residence by prefecture using the content of tweets and the observation weather data. From the experimental results, we confirmed that the proposed method achieved the F1-measure of 0.733, which is 1.6 points higher than the baseline model using only tweet contents.