Network modeling of a human-computer interaction task based on a self-organization method

A multi-modal interface system is proposed to realize an easy conversation between human and computer. The system uses a task model to infer a suitable response for multi-modal information. The model has network structure and is composed of relationships between word sequences and their responses. This paper proposes an inference method using a task model and an algorithm to self-organize the task model from word sequences and their responses using 6 basic rules and two types of background knowledge. We give an example task called "house designing" and show the selforganized task model is robust for noise.