미시건 대학교의 Applied Text Mining in Python 학습자 리뷰 및 피드백

Lectures are good but the assignment of week1 and week 4 is a little bit absurd and unclarified. Autograder is too slow.

Good for basic understanding only

Most assigmets were not in the notes. Still everyhting seems really usefull.

Graded assignments need more grounding in practically applicable situations.

I don't think the lecture is very clear. Although I have finished the homework, I still feel a little confused about the concepts.

Poor ability from lecturer to explain key concepts.

The course would be significantly improved if there were more hands-on demos during the lectures. Lectures are very high-level and aren't terribly useful when trying to do the lab exercises.

Good content.

Had to spend way too much time fighting the auto-grader.

good training

You do more work learning on your own to be able to do the projects and quizes then is given in the lectures. These University of Michigan classes aren't very balanced in terms of lectures, reading, and difficulty of projects.

This course did cover some good topics (Naive Bayes model, similarity, part of speech tagging). However, I felt the homework was more about manipulating Python data structures than learning anything significant about text mining. Some of the theory behind the models was covered, but didn't make it to the homework.

It would be difficult since this is a short class, but I would have preferred more about tips on which model to use and feature engineering / selection, and examples of practical applications of text mining. (Or stories of failures in the instructors' experience!)

The course is pitched at a introductory level. I would have like to have more practical tutorials.

The fact that the strategy of a Jupyter Notebook Demonstration during explanation was not followed in week 3 and 4 was a disappointment.This specialisation had been wonderful with its use of demonstration in Lectures with the Notebook,If this had been followed in Week 3 and Week 4 then the course would definitely had shined.Please correct the same, the course deserves that, It has wonderful content.

I think the course was superficial and could be better explored. It's good start, though.

I enjoyed this course, but some aspects of it felt "light touch", particularly week 4. That week would be greatly improved with a jupyter notebook and an applied demonstration by the absolutely awesome Teaching Assistant, Filip Jankovic. Whenever he does a demonstration, it's clear, concise, practical, and always helpful. Let's see more of him!

While the video of the course were OK, the assignments were of really bad quality. So many problems with the auto-grader, and some questions were absolutely not clear. I still put 3 stars because the subject is interesting and I got things I can work with out of it, but don't expect too much from it, you'll spend most of your time trying to deal with the weird assignments questions. For the time spent, they could have added 1 or 2 weeks of videos.

Course materials are amazing but there are not much support for assignments.

I did all the quizzes and assignments except the last one. It seems there was something wrong with auto-grader or the assignment was not clear. There were complains about this issue on the forum but no one from staff answered the questions.

Week 4 was not sufficient

The grading system is supremely messed up and at least I have a vague idea what am talking about because I have completed more than a dozen coursera courses. Also, the methods used through the courses teaches very bad coding approach relying on global variables.

Below is an example from Module 2:

def example_two():

return len(set(nltk.word_tokenize(moby_raw))) # or alternatively len(set(text1))

example_two()

Why would they not pass moby_raw and text1 as arguments in the function?

With that said, the course could easily be one of the best intro NLTK courses out there minus the frustration and very poor design.

Bad assignment. Grader not reliable. No control over the discussion board, many confusing comments mixed with good comments.

While the course gives a good broad understanding of how any NLP task would work in theory, but the course is very unstructured. For example, if I had to be a given task on doing a sentiment analysis, I can broadly tell what is the conceptual theory behind it but I dont know how exactly to do it because the professor talked about so many tools which were repetitive in their use and were not clearly demarcated as to what tool should be used for what?

I liked the course, but it felt as a very raw overview, I think it could have been more challenging when it comes to the models explained.

A bit lack of coherence in theory. Sometimes, the theory needed for the assignments was not given with enough detail, and you had to browse the forums for the information, and applying it to your assignment just to pass, sometimes without understanding why you were doing what you were doing.

More Python examples needed. For week 3, the tutorial about recommender systems was perfect for the assignment.

A lot of issues with the auto graders

I like the lecture very much. If the lecture can cover more example codes it iwll be greater.