| Budget Amount *help |
¥4,420,000 (Direct Cost: ¥3,400,000、Indirect Cost: ¥1,020,000)
Fiscal Year 2013: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2012: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2011: ¥2,730,000 (Direct Cost: ¥2,100,000、Indirect Cost: ¥630,000)
|
|---|
| Research Abstract |
There are more and more real-world deployed applications of embodied conversational agents (ECA's). These agent systems are usually used by groups of visitors rather than individuals. This situation is more complex than single user one, specific features are required. This project aims to build an information providing agent for collaborative decision making tasks. A Wizard-of-Oz (WOZ) experiment was conducted for collecting human/agent interaction data. By analyzing it, the model for identifying the addressee of user utterances have been developed with machine learning techniques on visual (face direction and movements) and acoustic (intensity, pitch, and speed) data of the corpus. We then incorporated it into a fully autonomous multiparty ECA. The system identifies the addressee and uses this information in language understanding and dialogue management. Finally, an evaluation experiment shows that the proposed addressee identification mechanism works well in a real-time system.
|
