Proposal of Summarization Support System Based on Mixed-Initiative Interaction

Abstract

This paper proposes a text summarization support system based on mixed-initiative interaction. Automatic summarization is used to generate sentences useful as a clue for information gathering. However, the quality of the generated summary is still lower than that generated by humans. Furthermore, it does not consider users' interest and knowledge. To solve the problems, the proposed system dynamically suggests what is useful for writing a summary, such as important sentences, in response to users' input. The effectiveness and challenges of the proposed approach are discussed based on the result of the experiment using a prototype system.