The community member is unsure if they can use llamaIndex with Dutch language documents. Another community member suggests that while OpenAI embedding models and LLMs generally work well with different languages, open-source models may struggle more. They recommend fine-tuning the models or using multilingual or language-specific models. They also suggest considering the prompts, as they are typically in English, and including prompts in the target language may improve response quality.
Yeah the OpenAI embedding models and LLMs are pretty good with different languages. For open source they will struggle more but you can finetune them like Elliot mentioned. You can also find some multilingual models or even language specific for some
You might also want to consider the prompts since they're English by default, with my language and 3.5 I found that including the prompts in my language helped with the response quality somewhat
