Add training codes for NER and format IDP4+ data to create baseline model

[rishabgit]

Download link

[rishabgit]

added the data prep codes with idp4+ download and conversion as default to set up the baseline model

[rishabgit]

@valearna Added the main training script, but I think it looks off compared to the pattern of other codes. It might be a good idea to rearrange this part later, so it's more intuitive to use

[valearna]

Ok, thanks @rishabgit. I'll also take a look at it

[rishabgit]

Refactored training script into an import like data prep

[valearna]

Can you give me an update on this part @rishabgit ? Is this all in the train_ner.py script? I could add an entry point to the script in the setup.py file so that it gets automatically installed when users install the genomic info package in their system: https://python-packaging.readthedocs.io/en/latest/command-line-scripts.html

[rishabgit]

@valearna It's a two-step process. You need to pull and process IDP4+ data, and then run train_ner.py script on it.

I added a readme with the commands - https://github.com/rishabgit/genomic-info-from-papers/blob/main/genomicinfo/entity_extraction/ner/readme.md

I'm not sure if running these two steps automatically when the user installs this package would a good idea, since then the downloading and training would happen by default for all users even if they plan on not using the ner block at all (TODO feature - #10). Also, in case user has an old CPU with no GPU, then training might take more than an hour (haven't tested exact how long).

[valearna]

Sounds good, thanks @rishabgit