Profile

Mike Turner profile image

Mike Turner

Student

  • 3 reviews
  • 2 completed
Content 
Instructor 
Provider 
I was originally excited to get a decent introduction to text retrieval methods in order to apply them in my current career, but this course falls very short of properly educating people on this subject and I would recommend skipping it. Here is a step-step review of it. 1 - The lectures The lectures and lecturer are actually decent, they introduce the concepts in a fairly clear way and cover a lot of good ground. I do not have an issue with the way this is presented in lecture. 2 - The assignments (poor quality) In this course there are quizzes and programming assignments, and the programming assignments are optional. This is not good, because programming assignments are how you learn. The programming assignments that do exist are of very poor quality. The first assignment is a step by step set of instructions to use a text retrieval toolkit (meta-toolkit.org) to do simple things, it doesn't make you program - only follow directions so there's little critical thinking involved. Secondly, it uses a very niche C++ NLP tool the instructor's grad student created called MeTA. It is not widely used and therefore the support & knowledgebase surrounding it is mostly non-existant and nobody representing the course answers the course forum, so if things go wrong, it is hard to find an answer as to why. Thirdly, currently the assignment files have not been updated in a long time, and they do not work correctly with the current version of MeTA (2.30) and thus you cannot complete the assignment without spending a lot of hours figuring out how to update the data provided YOURSELF to work correctly with the current version of MeTA. So the course's assignments are hard to complete due to negligence on the instructor's part. The second assignment is better, but also has recommendations out of date with the current version of MeTA. Overall I'd say this course suffers from 3 main things that make it not worth taking - It has you use a tool used by almost nobody created by the instructor's grad student that has zero support instead of better many supported tools out there. - Instructors & Course Representatives have clearly neglected the content & it doesn't work correctly forcing the student to fix the instructor's errors in the assignment content. - The programming challenges don't make you think too much, thus not giving you a very good hands on intro to the subject matter. There are better courses that teach this subject out there, and I'd recommend you skip this one in favor of one that uses better supported tools and better teaches hands on learning.
Content 
Instructor 
Provider 
I actually liked this class a lot. It is mainly Roger peng reading from slides, but he covers a lot of good ground in R and explains a lot of nuance. I had prior programming experience, but zero R experience going in and now after this course + the rest of Roger Peng's courses I am able to work in R. I think this course is best for those familiar with programming basics and no R experience.
Content 
Instructor 
Provider 
I passed this course with a %100 grade, but I might as well not have taken it at all. This course is TERRIBLY done. Firstly, from a delivery standpoint, it's horribly unpolished. The concepts are introduced in a rushed half-complete way, lectures often start and stop mid-sentence, the mathematical notation is incomplete and sometimes wrong, many of the SWIRL programming assignments throw errors and terminate halfway through the assignment, the class notes are very messy, and in some of the homework, you're asked to answer identical questions twice. As a lecturer, Brian Caffo wanders and stutters a lot and breezes through very incomplete explanations of crucial probability and statistical topics. When discussing how to implement statistical methods (t-tests, ANOVA pdf functions, probability distributions) in R, he doesn't really give an organized introduction to it, he simply plops code snippets into his lecture slides and stumbles through them without explaining much about what key R stats functions are and the nuances of using them. In the previous courses of the track that Roger Peng teaches, you get used to very well rounded explanation of R functionality. Do not expect the same in this course. Because of how poorly done this course was done, I've decided NOT to complete the rest of the John Hopkin's data science track. I recommend to the creators of this track that Brian Caffo be removed from this track and the statistical inference course completely reworked by another instructor.