Input/Output clarification

[amirj]

If I understood correctly, we have a dictionary which maps "aliases"/"synonyms" to a list of corresponding cuis.
The input of the algorithm is a string (mention) and the output is some items in the dictionary:

{
  "mention": "ataxia telangiectasia", 
  "predictions": [
    {"name": "ataxia telangiectasia", "id": "D001260|208900"}, 
    {"name": "ataxia telangiectasia syndrome", "id": "D001260|208900"}, 
    {"name": "ataxia telangiectasia variant", "id": "C566865"}, 
    {"name": "syndrome ataxia telangiectasia", "id": "D001260|208900"}, 
    {"name": "telangiectasia", "id": "D013684"}
  ]
}

Since our target is to map the mention to a CUI, I'm wondering if there is any functionality that map the above output to a single CUI?

[mjeensung]

Hi amirj

The predictions are sorted in the order of the final scores.
This means that the first item of the predictions is the top 1 prediction, and the CUI of it is the final single CUI for the given input mention.

[amirj]

Thanks Mujeen. What would happen if the top prediction is corresponding to multiple CUIs? --what's the best practices to extract the target CUI among them?

[mjeensung]

When a single prediction has multiple CUIs, the CUIs are from multiple KBs.
For example, in 'D001260|208900', 'D001260' is from MeSH and '208900' is from OMIM.

For evaluation, I consider it correct when any of the predicted CUIs is matched with the golden answer.
But, in practice, you can choose any CUI depending on the KB you are using.

[amirj]

Thanks for your clarification. In my use case, I'm leveraging UMLS "aliases/synonyms" in my dictionary. As a result, an ambiguous synonym would be mapped to more than one CUI.
What's your suggestion in this situation?

[mjeensung]

I think it depends on how you use it.
Is there a reason that you want to extract just one CUI?

[amirj]

Yes, I want to directly map a mention to only one entity, i.e. entity linking

[mjeensung]

Hi amirj

Thank you for your patience.

That's a good point. But in our work, we focus on handling term variations of the biomedical concepts rather than disambiguating mentions.