NebularGraph just support english? about nebula

[wey-gu]

So, I want know why my korean triplet translate english triplet
Does Nebula Graph have a function to translate the given text into English and save it?

This was due to the prompts inside the llama-index being all English-based, thus sometimes it confused what language to return.

I recommend highlighting this in the triplet extraction prompt to do so.

If still you cannot do it with the example, I can provide detailed examples.

Thanks.

[JinSeoung-Oh]

@wey-gu, Thanks! It is very helpful.
A few minutes ago, I re-run my code that has already failed dozens of times to return korean KG
And knowledgeGraphIndex return korean KG, suddenly....

Maybe your right. It is Llama_index error, maybe..

Thank for your help and have a nice day!

Thanks!!

[wey-gu]

Dear @JinSeoung-Oh , welcome to the NebulaGraph community!

We could override the prompt and (if openai was used, adding a system prompt to remind him to keep original languages may help, too).

In this example, I showed how to leverage models other than openai && how to override triplet extraction prompts for your reference(sorry it's in Chinese)!

Let us know your attempts and results!

Thanks!

[wey-gu]

Yes, I encountered this when building chinese kg with local models and the way to mitigate was to override the prompts for triplets extraction as in above example link :D.

Feel free to ping me for anything I could help.

PS. will bring more optimization for knowledge graph index/text2cypher/graph rag to llamaindex next week!