Natural Language Processing with PyTorch

I am new to NLP and deep learning frameworks, though I have advanced experience in structured data machine learning and basic experience coding my own baby neural nets from scratch in python. I was struggling to find a way into a deep learning framework like tensorflow or pytorch that would bridge the gap between my desire to take a particular problem formulation (inputs, activation functions, layers, output, loss function) and code it in a framework using best practice design patterns. Most framework tutorials I found start out more like a magic black box, and that's not the way I like to learn given my background. This book teaches NLP basics from the ground up along with a strong design pattern coded in python/pytorch. It teaches it seamlessly by starting from a simple example and continuing with other more advanced examples that keep using the same design pattern over and over again. For me, this is the best way to learn and remember. It has given me a foundation on how to sit down and code my own solution in an organized fashion using proper python object oriented practices. My biggest challenge moving from structured data problems to NLP is knowing how to represent the text input to a model and then feeling comfortable writing maintainable code that prepares the data according to that representation. The authors step you through this, and also even offer the code that munges the data into shape for each example, which I found helpful for rounding out my text-based python skills. I feel very confident I can build my own library or dive into other PyTorch/Fastai models knowing what is going on and what to look out for. So many helpful modeling and PyTorch details are given along the way. I'm truly grateful for this book to get me over the hump in the NLP space.

Pretty good introduction book for NLP newcomers like me. It kinda goes all the way from the beginning (perceptron and basic pytorch) so if you have never been exposed to NLP this is the right book for you

The book is thin, but it's concise. They not only cover the basics of NLP, which I needed to clarify a lot of my previous studies, but they give complete examples in their GitHub repo that just work with minimal fuss/muss. They get you going with PyTorch from the ground up. Everything is absolutely clear (to me) and there's no "clearly" or "left to the reader" academic bs. Some disclaimers: you should at least be familiar with numpy/pandas and have a solid grounding in Python. Yes, I will finish this book and it's examples in 2-3 weeks, but it has truly jump started my efforts in a new domain.

This book sets out to serve as a text for those new to deep learning and it does just that. For those removed from research, the mathematics of other introductory texts can be a bit distracting. This text takes a refreshing step back to show what you might see in the research as code. It runs the gamut from training your own models from scratch to transfer learning in the course of a few hundred pages -- what a feat! As an editorial note, I find the bitly notes a bit noisy -- I might have used the referent links in their place, but no big gripe. Bravo, Delip and Brian!

The book is well written and provides some valuable insights about the "why" we need the different NLP models and "how" they differ one from the other. However, it is quite thin and some topics are not covered in detail. More importantly, 1-2 or more pages codes are presented with brief (only a few lines) explanations. I would prefer instead simpler problems with shorter codes that can be explained in detail. Finally, the area of NLP has evolved a lot during the last 2-3 years (after this book was written) where attention mechanisms, new embeddings and transformer architectures have now become the new state of the art. Thus, some of the newest content is missing from this book.

While this book covers some interesting topics, it is very code heavy and the code it provides has a lot of issues. First off, the code is poorly structured. If you try and follow along with the exercises, you will end up typing the same boilerplate code over and over. Second, really important pieces of code are missing from the book. The book's code will call functions that are not references any where else in the book. The only way to find out what they are supposed to do it to hunt down the functions in the books github repo. Third, there are several code examples that just don't work. If you try to run them, you will end up spending a lot of time debugging code that is just wrong. I want to like this book, but if it is going to focus so much on code, it needs to have higher quality standards.

An awesome book to dive into your NLP and deep learning journey with PyTorch. Delip and Brian have done a great job in explaining NLP concepts clearly and demonstrating them in code in each chapter to solve practical NLP tasks. Disclaimer: I was one of the technical reviewer of the book =)

As a 15 yr veteran of NLP/text mining research, I was disappointed at the lack of original and timely content. Quite a lot of review of my credibly basic stuff, and not much more timely than the earliest embedding schemes.

Natural Language Processing with PyTorch

I was probably the first customer from India to purchase this book. Tweeted the book to Delip (Author) when it arrived. I agree the book doesn't have advanced concepts like transformers (that is indispensable for NLP today). But NLP at the very basic level is about understanding and being able to build from scratch. I think this was the goal of the book and it achieved that goal. it is better than high level like FastAi if you want to really want to understand what your code is actually doing. The Author also has provided a GitHub repository of all the examples in the book that was quite helpful. In short, if you want to build a solid foundation in NLP, read the code and absorb this book. The materials are not sufficient but they are necessary for one to have a solid NLP background.

I'm disappointed already on page 8. Axis labels are blurry and barely readable. It is a black/white copy although the book is in color originally. The very first code example on page 7 is broken. The book is overpriced given the quality of an average blog post. Sorry, didn't read any further.

Kindle Qualität ist sehr schlecht. Formeln werden nicht richtig dargestellt. Bilder / Graphen abgeschnitten

The book lacks proper explanation. I don't mind paying 1500 bucks for a good book. Bit in this case it is pricy.