Make running the script a bit more simpler.

[hunterhector]

We should consider reducing the difficulties of running the example codes, including making "model_download" and data download automatically, and placing the core instructions in a more clear place.

Here is a good example:

https://github.com/asyml/texar-pytorch/tree/master/examples/sentence_classifier

[jzpang]

"model_download" is done in requirements.txt. For data, it was provided in both Kaggle and semanticscholar, Kaggle requires log in so we could not provide download link, and for semanticscholar, I am not sure if we could set up automatic download because of license issue as it claims

'By downloading this dataset you are agreeing to the Open Covid Pledge compatible Dataset License for the CORD-19 dataset that details the terms and conditions under which partner data and content is being made available.'

Any ideas?

[DianLiI]

https://ai2-semanticscholar-cord-19.s3-us-west-2.amazonaws.com/historical_releases.html
Download links here do not need credentials.

[jzpang]

I also found this link in download of this sementicscholar page https://www.semanticscholar.org/cord19/download, which claims the licence thing, is it okay just to include the download link in the script?

[hunterhector]

We can include a claim and a copy of the license in the readme file.