AskML: Pointers towards using appropriate NLP techniques in a ML problem

I want to use Natural Language Processing techniques in a Machine Learning problem.

The problem is to find how many upvotes a new comment in a website like reddit/Hacker News might get. Especially on the posts submitted in r/new. Other than the comment itself, I am also provided with meta-information about the user's comment karma, posts karma, her usual activity hours and how long she has been registered.

My background is in Operations research, I have taken classes specializing in Statistics, Machine Learning, Optimization Theory. But I have no idea about text processing. I want to know where to start. And what would be appropriate techniques to quantify the usefulness of a comment. As much as I would like to learn NLP in depth, possible quick fixes will be really appreciated.