Automatic generation of dominant patents in blank spaces: tfidf/mbart/mt5/CLIP embeddings cluster reconstructing vis: a means to visualize, quickly browse, analyze, and search for dynamic text-to-sentence similarities of patents and other long texts. and a third patent search method, a dynamic knowledge extraction management method, and automatic patent generation.

I use this tool for kaggle and other data analysis competitions.

*This code has a 3-year track record. After refactoring, we plan to release it to the public.

https://qiita.com/kzuzuo/items/8a80d8974bf3a7db7e54