Toward an outsider agent for supporting a brainstorming session --- an information retrieval method from a different viewpoint


Conflicts among different concepts are often useful in creating new ideas. Therefore, an outsider's attendance at a brainstorming session is often effective for activating a brainwave. Our research goal is to construct an artificial outsider agent. As the first step toward the goal, we propose an outsider model as an information retrieval model for obtaining information that has not only evident relevance but also hidden relevance for users. A prototype system has been constructed. Subjective experiments using the prototype system and a detailed analysis of the results confirm that the outsider model can extract such information. Furthermore, we propose a method to combine plural domains of knowledge. It is expected that the combining method can generate new relevance that could not be obtained by using the knowledge domains individually.