Manning and Jurafsky: Why this course doesn't seem reasonable for first is that it talks about more about writing algorithms. Tweets and poster are publishing the infos. Some basic people joining the course:80,000 people join, 4000 people regularyly submit tasks and codes.
The standard to apply algorithms and how to evaluate the NLP systems.
training set test set hidden test sets development set
What you can do: A Sentiment classifier. The aim: important features.
Why the NLP systems want to share. Why over-fitting is not suggested.