Human-Centered AI
2019-06-01 – 2022-02-28 · Completed
National Research Foundation of Korea (NRF)
Results
- Developed FAQ recommendation using learned sentence meaning and similarity.
- Converted agents’ final answer selections into retraining data.
- Implemented a web application, API and consultation-support prototype.
Evaluation results reported in the 2022 completion report.
| Metric | Result |
|---|---|
| FAQ collection | 2,808 entries |
| Top-three accuracy | 94% |
| Throughput | 33 queries per second |