Participation in open source coding programs
the-ethan-hunt opened this issue · comments
These days, a lot of programs have been initiated by technical organizations to get students acquainted with open source. One such program is the Kharagpur Winter of Code held by KOSS, IIT Kharagpur in the month of December since last two years. I understand that awesome-NLP is a curated list but I believe we can benefit greatly from this program. I would like to hear thoughts about this from @NirantK and @keon . In case there is no available time, I can act as a mentor. 😄
I am actually working on a small project where we release code samples. If awesome-nlp is a curated list, that is more Curated List of NLP Recipes.
It covers some of the concepts and the libraries mentioned in Awesome-NLP as Jupyter notebooks. Do you think KOSS might help coverage there?
If yes, we can simply merge the two projects to awesome-nlp by adding a directory to this repo. Leaving the list undisturbed.
Yes, that sounds great. The program invites both beginners and expert open source developers and I find this project good for someone who is good at NLP and open source. KWoC is a program on almost an international scale and I am pretty sure KOSS will get us coverage for it.
Sure, sounds great. I can help with what we want to do. It's basically for two tracks:
- Recipes: Most common paradigms e.g. embed attend encode predict and approaches in non-Python, non-English languages.
- Evaluations: Setup automated evaluation (similar to Kaggle) for NLP tasks ranging from text classification to coreference resolution. This would become a global leaderboard on those selected tasks.
Didn't get the evaluation idea properly: do you mean to setup a leaderboard with submissions on NLP tasks like Kaggle does on data science?
Yes, that'd be a good way to put it.
In fairness, I don't think Evaluation is a good fit for awesome-nlp . We should rather focus on Recipes here.
Alright. But what if we don't abandon Evaluation altogether? How about we set up another repo for this?
If we are going to use another repo, might as well work with NLP Progress which maintains a leaderboard on a wide range of NLP Tasks. What do you think?
Sounds good. But we need to ask @sebastianruder for that
Sure, I don't have the time to take Evaluations forward right now. You are free to reach out and coordinate with him.
Great. So should we submit awesome-nlp Recipes as a project? Or an independent repo?
That sounds exciting. You can include a link to a leaderboard on NLP Progress as long as the dataset is publicly available. Ping me again once you have something and we can see how it fits.
Let's incorporate all the ideas we have on the project:
- A new project on
awesome-NLPforRecipes. This includes NLP approaches in Non-English languages - A new repository will be created for
Evaluationsor work from NLP Progress itself. I can take it forward and mentor it. If working in an independent repo, after the program if @sebastianruder permits it, we can include it in NLP Progress
Please correct me if I am wrong anywhere.
That sounds good @the-ethan-hunt. Let's focus on either one for the time being. Strictly from the point of view of utility, I think recipes are more useful to beginners and evaluations to practitioners.
Thanks for the support @sebastianruder - will discuss if/when we do work on it. If you have an opinion of what might be more useful to you (and academic more broadly) - that'd be great.
@the-ethan-hunt - would you be taking the lead on this one?
If yes, it'd be great if we could sync over email or here and plan this better.
Sure, I will be happy to. 😄
Closing this for inactivity. @the-ethan-hunt please do share any updates that you might have for us :)