stanford-oval / WikiChat

WikiChat stops the hallucination of large language models by retrieving data from Wikipedia.

Home Page:https://wikichat.genie.stanford.edu

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Integrating Wiki Knowledge from Multimodal Short Videos

ScarletPan opened this issue · comments

Fantastic work! Have you considered incorporating knowledge from multimodal short videos into this project?

We have developed a multimodal short-video encyclopedia that links videos to Wikipedia items and aspects. It would be amazing if you could leverage this data to address "how-to" knowledge hallucination challenges:

GitHub Repository: https://github.com/KwaiKEG/Kuaipedia
Dataset: https://huggingface.co/datasets/kwaikeg/Kuaipedia

Thank you for sharing your work, seems quite interesting!
I took a quick look, and I think the dataset is only available in the Chinese language, linking to zh.wikipedia.org?
This would mean we need to extend WikiChat to Chinese first, since right now it has mainly been tested on English, and uses the English Wikipedia as its knowledge source.

I would be happy to hear your thoughts on this, if you are intersted.