Speakers
Description
Students often need good practice not only for English oral presentation but also for oral answering in Q&A sessions, especially when they present their research results for the first time at international conferences. In this study, we aimed to assist those students in improving their oral skills in Q&A by having ChatGPT read the papers, which were written by the students, and ask general and specific questions on the papers to the students. We examined six configurations of ChaGPT for question generation to find the optimal one for this task. The six configurations are a combination of whether reference papers should be explicitly uploaded to ChatGPT or not, and whether questions should be generated for each keyword, for each section, or not specifying any scope of the paper. We recruited eight master students who recently attended their first conferences and made oral presentations there and eight Japanese teachers of English. Each student evaluated the questions of his/her paper in terms of relevance, specificity, clarity, inspiration, and expectedness, but the teachers evaluated all the 471 questions paying attention to whether the questions were likely to be raised by language teachers or only by specialists. Finally, all the students and teachers selected the best configuration for Q&A practice. It was very unexpected to us but their responses indicated that the simplest configuration, where reference papers were NOT uploaded explicitly and any scope was NOT specified, was selected as the best one. Following this result, ChatGPT in the simplest configuration was integrated with a speech interface, ASR and TTS, so that ChatGPT can ask orally and students can answer orally. Although the integrated system has not been evaluated yet, we believe that it can help students prepare for their first encounter with international academia.
| Keywords | ChatGPT, Q&A, oral skills, automatic question generation |
|---|
