学生对 deeplearning.ai 提供的 Natural Language Processing in TensorFlow 的评价和反馈

Good course

good

Very disappointed in this course. Instructor seems to have limited understanding of how sequence models and word embeddings work, or is unable to communicate the ideas in his teaching. Explanation for the theory is limited, and he has difficulty tying theory to the TensorFlow framework. Not sure why you would begin teaching sequence models with LSTM blocks combined with standard NN, way too complex structure. Instructor doesn't talk about why sequence models are important and useful in the first place. Very very poor.

Instead of taking this course, I could've read the tutorials on the TensorFlow site. Those tutorials are regularly updated, maintained, much more detailed and they're FREE.

This course, along the other courses in this specialization are not good for other than exposition to the TF API. Actually, they're not even good at that because the TF tutorials do a much better job at that.

And it's so frustrating that over-fitting is never tackled in any way and not even a hint at how to solve it is even given.

This was by far the hardest course on the sequence. I actually skip it and did courses on order 1, 2, 4 and now 3.

* Notebooks were not as easy to follow. Maybe put more comments on what was expected and describe the datasets a little more.

* There are typos here and there, for instance "The pervious video referred to a colab environment you can practice one. "-> previous.

The file at https://github.com/tensorflow/datasets/blob/master/docs/datasets.md NOT FOUND

The course is merely a walk-through some Jupiter notebooks of Laurence. There are no proper slides with explanation of what's going on. I also don't see much activity from the course creators on the discussion forums. It is incredibly easy to complete the course without forming any deep understanding.

The weekly programming exercises are not even automatically checked for accuracy.

I am taking these courses to learn via example. (this is not theory course, it is a course on practice). The fact that there are not well thought out programming exercises makes this course much weaker than the proceeding two. The first two courses in this series are much better for this reason. This course looks unfinished. The lectures are okay, the quizzes are okay.

Very limited practice examples for learners. Also the example are very simple. The course should have been made much detailed and much real example problems. For instance, in the Week 4, topic 'Text Generation', generating a Shakespeare poem seemed to be a very silly example. The quality of Coursera Courses are becoming very poor.

The material was better in this course than the previous ones, but still lacking depth in my opinion. Also, no graded assignments?? So the focus is then only on the quizzes, and they are not even well done. From week to week the same questions are repeated and the quizzes don't even include code: How is this teaching code?

This course was quite sloppily presented and superficial overall. There were a couple of longstanding errors that have never been fixed (see the lengthy discussions in forums). One thing that annoyed me was that the important concept of stop-words was not discussed at all, yet it was required for the first assignment.

All assignments were optional - probably because of all the problems with the scoring system for the previous course. Quizzes often asked things about the dataset we used (eg IMDB) rather than testing that we were learning concepts. Very little meat to the course - mostly links to other resources.

After completing the Deep Learning specialization, which I really liked, I had high expectations for this one. Unfortunately it can not meet my expectations and was a dissapointment. Even if I try to see it objectively and ignore my high expectations, the quality of the exercises were very poor.

Overall the video material is fine. The assignments however are very unclear and contain bugs. The grader's test don't match the instructions. It's very frustrating that the assignments clearly haven't been given the same attention the rest of the course has been.

This course is a far cry from Andrew Ng's deep learning specialization and refers to Sequence Models from that specialization at the drop of a hat. In short, no use doing this one, unless you have done sequence models (course 5) of deep learning specialization.

No grader exercises,

sample code in the lectures isn't always updated and gives errors,

everything is a bit chaotic (eg order of sample code, sample code description, introduction to the topic is random; some random parts in the code).

The weekly exercises are not graded and the over all content quality of this course in comparison with the previous two in the specialization seems a bit poor and doesn't provide more learning on the topic.

Not clearly explained and only using toy and irrelevant datasets, nothing realtime industry specific examples. Also, voice quality is very bad for this course.

I would call this entry/beginner level material. There arent any graded coding challenges, which is a shame. No complex topics are covered with this class.

It's a bit too basic and there are not many graded examples to work through like Andrew Ng's course. I feel it could have been more complete and in depth

Disappointing. Laurence much less able to explain NLP issues than CNN issues. Lots of problems with TF versions in Colabs wasted far too much time.

The quality of the audio recording is worse than courses before. The questions at the end of the chapters are also repetitive.

The given material is pretty nice, but I don't feel like I learned a lot. Important concepts weren't explained in depth.

didn't really feel like a strongly grasped the concept and needed more exercises also the lack of lessons notebooks.

Too short. Fine as introduction but not in depth course. No assignment except very shallow multiple choices tests.

some links are broken.

The last quiz (week4) not match to the lessons

couldn't run the lab excercizes