jgc128 / mednli Goto Github PK

Hi

What is performance of the pre-trained model? is it the same model which achieve 73.5 accuracy on test dataset (Table 3 in Romanov and Shivade, EMNLP18 paper)? It will be helpful if you add the performance in the table in the below screenshot.

Also, when do you think you can share the more pre-trained models?

Thanks :)

I do have a membership to physionet but still can not get data
https://physionet.org/works/MIMICIIIDerivedDataRepository/files/approved/mednli/

Hi, I've completed the course and uploaded the training report. Could you please help me review it? I'm a bit anxious to get this dataset. Thank you very much.

I am following the instuctions to download the dataset, but when I tried to go through the CITI training, I found the CITI website was different from the step-by-step instructions: https://physionet.org/about/citi-course/ . And I don't know how to learn the course. Can the instuctions be updated? Or there is another way to get access to the datasest?

Thank you for providing medmli! I was wondering if you could assist me by telling me how could I run your code like InferSent demo.pynb?

https://github.com/DAVEZJ/Simple_Sentence_Simliarity/blob/0f564adedad62f1bbe13265ca3848d98e8b363f6/Simple_Sentence_Similarity.ipynb

Hi Alexey -- Sent you an email re: this but just wanted to post here as well in case this is a better way to reach you. It looks like the MedNLI dataset has not been migrated to the updated physionet.org website (it is still located on archive.physionet.org, the old version); although we have an official physionet login we cannot access the raw data. Would love to use it to train a model I've been working on.

Please let me know the best way for me to access the MedNLI dataset. Thank you!

Thanks for providing the pretrained embedding models! I know it's been a while since you have exported these models. Can I please ask if the provided Fasttext models contain subwords at all? I have converted them to the format that can be imported using Gensim but found there is no subword information. Am I right to assume the subwords were not saved as part of the models?