つぶやきロボット : インターネット資源を利用したユーザの要求発見

ロボット工学

日本知能情報ファジィ学会

ファジィシステムシンポジウム講演論文集

In this study, we describe using text information from Twitter to understand context for task selection of a robot. An intelligent robot of the future is required to discover users' discomfort and im-prove it whenever possible. However, the level of image or speech recognition is far from satisfying, which does not help robots to understand users' needs from context when an order is not explicit. Therefore, we decided to utilize the text information to automatically discover situations when robot's help is needed. Users share their information about their daily matters on Twitter. A vacuum-cleaning robot acquires users' "tweets", and starts cleaning if it is recognized as a description of situation requir-ing action. Additionally robot can reply user when there is an explicit request but the robot is not able to perform the task. Moreover, a chat module is triggered on to answer tweets which were recognized as unrelated to the cleaning task.