Data Science Capstone

The positive: Completion of this project requires most of the skills you will have learned in completing the prequisite courses.

The negative: If you're "like me" (inexperienced with NLP), you should start reading up on the basics (enough to know the challenges, available tools, prevalent techniques) 1-2 weeks before Week 1 of the course. If I hadn't done this, then I would not have given myself enough foundational understanding to understand what work(s) I should use as as my guiding examples (and why). Reading the course forums and reviews, I think many people who had trouble completing the project were not able to take sufficient time before that first week to get oriented.

Suggestion to improve: Give people a heads up, as well as a list of digestable reading material and suggest they start reading 2 weeks before Week 1 of the course.

Details:

The domain of subject matter central to the capstone project (text mining and NLP) differs from that of examples used in the prerequisite courses, but the prerequisite courses gave me the tools to tackle the new domain. Now, I did have to read up on text mining and NLP. I'm a slow reader, so it took me a couple of weeks to read enough that I felt comfortable with the concepts that would help me complete this project. Once I took the time to get that background, though, it became apparent that "predicting the next word" is an NLP challenge that has been thoroughly studied and there are LOTS of examples freely available, where people describe and show how they have tackled the same challenge.

Since I have a software development background and completed the prerequisite courses in the specialization, I was able to translate approaches described in English into software. I was helped along greatly by the existence of R packages such as tm and tokenizers, and was always able to find guidance on addressing thorny issues by using "good ole Google Search". Most often, my searches would lead me to StackOverflow and writeups from capstone project alumni. I did my own writeups and wrote my own code, but I also very much learned lessons from those who went before me.

I would recommend this specialization by JHSU, which (as it should be) is a package deal with the capstone project. Applying what I learned to a new domain helped me better understand how to practically apply sampling, statistical inference, cross-validation and more (just to another domain).