Text Generation in an Interview-Style with GPT-3


I am a beginner in Text Generation and wanted to get some advice on how I could go on with my project.

For my project, I interviewed persons with the same set of questions (one control group, one target group) and then fine-tuned a text classification model to predict if a person belongs to the control- or the target group, based on the answer of the specific questions.
Now I am lacking data so I wanted to synthetically create some answers for both groups. My plan was to use GPT-3, and feed it with my data so the model is able to learn the style of the different answers.

Is it possible to synthetically create answers in an interview style with GPT-3?

Thanks in advance for any advise!


Did you ever find a solution to this? I am also trying to train the meta llama 2 model (or any model) to be fine-tuned with transcribed interviews held inside zoom. I don’t know how to proceed yet.

Any help would be greatly appreciated. Thank you!