Comment by shchegrikovich

I discussed this a few weeks back. The idea is to take a Python dataset, as Python is the most popular language, and write a transpiler to Prolog with the help of llms. So, creating this synthetic dataset is not a huge problem.