Along with the “W8.1” video on Overfitting, I’d like you to watch the video below. It is more or less a different explanation of the same ideas, and I think it would be useful for you to watch both:
https://www.youtube.com/watch?v=EuBBz3bI-aA
The jupyter notebooks that I used, as well as the paper that I show in the last video of this week, are available in the Download Folder.
Before proceeding to the next video, I’d like you to watch these two videos from Youtube. I believe you should be able to understand them well with the information you have so far. Let me know if you have doubts =)
I am not really sure if these videos “should” be in Youtube (because it looks like Youtube keeps shutting them down, and then that channel just reuploads them a little later – but in the end this means I keep having to update the link to them here). In any case, if the links are broken, just look for their in Youtube.
Title of the video: An Intro to Parts of Speech and POS Tagging https://youtu.be/tJBvmkNsoN8
Title of the video: Some Methods and Results on Sequence Models for POS Tagging https://www.youtube.com/watch?v=QMZsurbHVwQ
Additionally, if you still didn’t watch those 3Blue1Brown videos on Neural Networks (the two videos on classifying handwritten digits using a neural network), I’d like you to stop now, go back to the last week’s materials, and watch it. You will need that information for understanding the last video of this week.
The following video is about a few common NLP tasks. POS Tagging is normally solved with statistical models (like Neural Networks); but the other tasks I refer to in the video are normally not really “learnt” statistically:
In the last video of this week I “dissect” the following paper:
However, since I go through the whole paper in the video, explaining what it does and showing its whole text, it is probably not ok for me to upload the video here.
(I’ll see what I can do. Just reading the paper is not going to be super interesting: the reason I took this paper is because I could show how the few concepts we’ve learnt already are enough to understand quite a lot; but it will be hard to map the concepts we’ve learnt without some guidance =/ )