Data is not downloadable

[Asad Mahmood]

I tried running the piece of code given in the ReadMe.md but am running into an error which is related to data folder. I wasn't able to download data.zip as it takes me to a 404 page. Can you guide me?

Thanks.

[Dan Lou]

Hi,

Sorry for the late reply... Upon closer reading of the terms, I've confirmed that I'm not allowed to share data derivative from UMLS, unfortunately.

Still, as I pointed out in the README, you should be able to re-generate these contents using the create_umls_kb.py script.

As another alternative, you may be able to use/adapt scispacy's UMLS KB as a replacement.

In case you're only interested in the adaptations to the MedMentions dataset, I've uploaded that separately here:
https://drive.google.com/file/d/1wJdW3Tcb6VZ0z-d8XQahk2Gm4Cj0BrRu/view?usp=sharing

Best

[Keshav Biyani]

Hello, even I am facing a similar issue. I tried to run the create_umls_kb.py script but it's giving the following error:

"""

Traceback (most recent call last):
File "scripts/create_umls_kb.py", line 10, in
umls_tree = construct_umls_tree_from_tsv('data/umls_semantic_type_tree.tsv') # change to your location
File "/home/keshav/anaconda3/envs/medlinker/lib/python3.6/site-packages/scispacy/umls_semantic_type_tree.py", line 82, in construct_umls_tree_from_tsv
for line in open(filepath, "r"):
FileNotFoundError: [Errno 2] No such file or directory: 'data/umls_semantic_type_tree.tsv'

"""

I tried to download the data from the google drive link. I have requested for the access as well.
But still there's no luck. Can you please let me know what do I have to so ?
Thanks
Regards

[Dan Lou]

Sorry, didn't realize that had restricted permissions.
I've now accepted your request and updated permissions.

Best

[Dan Lou]

In case you're having trouble accessing the umls_semantic_type_tree.tsv file from scispacy, you may also find that here:
https://drive.google.com/file/d/1UGRWvynFmLb5gSF0kc16Bsh4DTCdVMJ2/view?usp=sharing

[Keshav Biyani]

Hello Danlou
Wished to ask one more thing, how do we train from scratch ?

[Dan Lou]

The code available in this repo can help you train from scratch.

Check the 'create' methods in the 'matcher' scripts, as well as precompute_contextual.py for extracting embeddings from the NLMs.