Abstract
With diffusion and development of information-related technologies in recent years, geographic space information services, typified by car navigation, are widespread. However, provision of map information, which is the basis of geographic space information services, depends on on-site investigations conducted manually; therefore, there is a problem that the map information cannot catch up with the changes in the real-space information, such as the opening of a new store. Therefore, the authors have been conducting a study on text mining to automatically collect store information such as store names, addresses, and the industry types from the natural language, with focus on the information about the opening of new stores on the Web. However, when conducting the text mining using only the keywords related to new store opening, we were faced with two problems. One is that the differences in the word appearance tendency by industry type or the combination of words before and after one another become a noise, which causes the accuracy of the collected store information to fall. Another problem is that since the precision of the information sources that may contain typing errors or false information is not guaranteed, it is difficult to judge whether the data is correct or not if there is inconsistency in the contents. This study aims to overcome these problems by mining using the keywords, and by considering the fact that there is a strong correlation between the industry type of a store and the geographic characteristics of the store location. By experimenting, we verified that this study considering the geographic characteristics of the store location was effective to collect the new store information.
