[Question]: Could Keyword component support custom keywords?

[yangxg]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (Language Policy).
• Non-english title submitions will be closed directly ( 非英文标题的提交将会被直接关闭 ) (Language Policy).
• Please do not modify this template :) and fill in all the required fields.

Describe your problem

I am using the keywords component to retrieve the knowledge. It seems better than semantic similarity approach in my case. I also noticed that ragflow are primarily using keyword approach for knowledge retrieval (the link).

However, when I query with some domain keywords, the exist keyword component only gives the conventional result, and the subsequent knowledge retrieval maybe not correct.

For example. in my knowledge base, there are chunks on both "全市三级医院床位数" and "全市二三级医院床位数". When I query "某市三级医院床位数"，the keyword component usually returns “全市 三级医院 床位数”. But when this result was insert into the knowledge retrieval, it often gives the indicators on "全市二三级医院床位数".

I suppose if the keyword component could support custom keyword, and if I could sent the custom keywords like "全市三级医院床位数" into knowledge retrival component, it should return much more precise result as expected. Thanks.

[KevinHuSh]

I don't quite follow.
Do you mean Retrieval component in Agent?
Or, this may help.

[yangxg]

Hi @KevinHuSh . I am sorry for my confusing post, let me put it simple.

What I expect is mainly knowledge retrieval (either in Agent or in Chat). Following this an screenshot of a sample knowledge base, parsed by table approach. And there are records on specific indicators.

When I try to make some search, for example "各医院四级手术占比情况" , the query string seems got segmented.

Although the result is correct, I am wondering if we can use some custom dict to get more precise result? For example, using the specific term like "四级手术占比" to make the Retrieval. I mannually make a screenshot just for illustration.

Thanks for your consideration.

[KevinHuSh]

You could test your query in Retrieval test.

[yangxg]

Thanks,I have tried retrieval test, the result content is segemented. I suppose there is a way to put custom dictionary in the segementation tools.

I have found the rag_tokenizer tool, it seems have the function to load userdict (loadUserDict(self, fnm)). Could you give some hit how to use this function?

[KevinHuSh]

Add your terms to ragflow/rag/res/huqie.txt in the format like:

TERM 1024 nr

DO NOT forget to delete ragflow/rag/res/huqie.txt.trie after altering huqie.txt.

[yangxg]

got it, thanks!