5 key takeaways for NLP course from High School of Economics
Preface: Course is a part of Advanced Machine Learning specialization on Coursera and is one the most rated NLP courses existing on that portal. Link to course: https://www.coursera.org/learn/language-processing/

Second part of 2019 for me is running under theme of self-education in the area of Machine Learning and related financial investments are inevitable. My membership on medium.com is without a doubt the most efficient investment yet as only for 5$ I’ve saved and read (almost) 300 blog posts relevant to the subject and countless others. Second to medium is subscription on Coursera. When I was looking for a course to start with on that site, my selection approach was as simple as possible — take the most rated one. And somehow it did workout…
1. Course is not over-rated
First and foremost, is it actually 4.6 out 5? Well, yes. From the perspective of my “not-so-huge” experience of online learning I can tell that it’s one of the best courses I’ve been investing time, money or both. You’ll get a good grasp on core language analysis problems and review of most common solutions for them. Be prepared to see a lot of math as authors heavily rely on it during presentations. While at the beginning it might feel a bit annoying, closer to end it was actually the twisting point for me, after which I’ve started to actually read scientific papers instead of just bookmarking them. If you are like me one month ago and you fell deeply uncomfortable with all that mathematically flooded pdf’s from arxiv — try this course. You might not eliminate the lack of comfort but you will at least know how to approach such texts.
2. Invest time in practical parts of tests
…And the exact mechanism of understanding the cumbersome math formulas is to apply them on some toy example that in most cases would be kindly provided to you by teachers. While every quiz in this course contains at least one calculation-based question, sometimes due to grade threshold you don’t need to solve all tasks to complete quiz. Don’t use that as an excuse to skip such parts as it will negatively influence your learning experience.
3. Forum is a part of learning material
Discussion section for this course is structured for each week + some additional topics to discuss tech related questions or to talk with teachers/TA’s. For my on experience starting from week 3 you will benefit from reading at least few topics from your classmates and maybe even share your own struggles or achievements. Specially important it might be during last week, when you’d have to tackle a lot of non-ML stuff to complete assignment. Don’t be shy and enjoy community help!
4. Reviews are crucial part of learning material
To be honest, 50% of my learning experience of this course was gained by practical assignments and review of classmates projects. Working through reviews you’ll have to complete specially prepared checkbox lists and they are there not only to help you with grading others — they will trigger you to grade yourself. You will also have a chance to do something that’s rarely available for people not directly involved in producing ML-projects for money — code and deal with bugs. Don’t get me wrong, we all complete placeholder materials from books or copy solutions from medium posts in order to play with new interesting algorithm, but that’s different. Even if you manage to spoil something in initial solution, very likely that it would be noticeable soon and you’ll not have to deal with consequences of your mistake. In this course you’ll often find yourself in a situation, when several inaccurate key prints made your code not ready for submit. Literally, just one mistake in dimensions length made my week 3 assignment so huge in terms of data amount, that coursera refused to except it via http :|
5. You can make it in one month
For those of you who are trying to estimate the cost of such investment, I can promise you that with spending at least 8 hours a week you’ll be able to complete this course withing one month and pay only for it. Try to be very fast at first weeks as they offer rather simple tasks compared to second part of the course. Myself I was able to complete weeks 1–2 within six days. After that use gained time for deeper work on weeks 3–5 as you’d likely to get stuck on something (damn you “rank_candidates” function, I’ve spent two days on this small python snippet!)
That’s it. Hope I’ve made you at least interested about this course. If some of you already passed it or expect to pass it in the future, feel free to reach me out using comment section with any questions or feedback you’d like to share.
Thanks for attention!