Thankyou for starting #25DaysOfNLP @abhi1thakur .
Here are few things that would be nice if covered to some extent . Would love to contribute as well. Please fell free to add if I miss something.

#nlproc #MachineLearning

NLP Introduction : What is NLP and why should we care about it. Would be nice if we can discuss Classical NLP vs Advanced NLP using Deep Learning. Should we care about both ? If yes, why . This could be a good start for beginners.

There are so many amazing libraries out there. How to pick one? Now this may sound simple to someone who has some experience with NLP but someone new to this field can get confused preety fast. @huggingface @spacy_io @ai2_allennlp @h2oai @gensim_py @fastdotai @NLTK_org

Wow now we have answered fundamental questions we can really jump quickly to the basic building blocks of NLP. This may cover range of topics including cleaning textual datasets, vocabulary, tokenization etc.

Now we have learned the basics and also learned the tools at hand. It's really nice to start implementing something rather than just going on and on to read ton of information as it may never end.

Github : https://github.com/Speech-and-NLP/awesome-project-ideas * Work in progress *

Now there are such a huge number of tasks that you can achieve using NLP. Just pick a task that you would really love to solve and bring something out of it.

NLP Progress by @seb_ruder : https://github.com/sebastianruder/NLP-progress

One other skill that would be really really useful here is how to keep yourself updated with NLP. This field moves really fast and this is one thing you just cannot ignore. If you have to take away one thing from this whole thread try to pick this one.

Let's now focus on representation of textual data. How to represent words, characters, sentences or documents. Which is better for your project and why to choose one over another?

Next one of the crucial thing to decide is what architecture to use. Now transformers is used mostly in the field now a days but try to use very classical and interpretable models at first. Now how to decide which model to use ?

Now we are really picking up pace. I remember the time when this blog was published "NLP's ImageNet moment has arrived" in 2018 and it really changed the pace of the industry. Learn what really fueled this growth.

Blog: https://ruder.io/nlp-imagenet/ 

Now we have seen a lot. But let's think that you decided to build an sentiment analysis model. How to build an state of the art model. Here come the @huggingface in picture.

Models : https://huggingface.co/transformers/model_summary.html

Damn

Let's take a look at NLP conferences. As what if you have built something amazing and want to share with the world or maybe start your next startup.

This was like brief idea of what we can do this was more for people who are starting out on this journey rather than an in-depth guide. Happy Learning