First, start with an idea of what you want to look into.

Find at least one really good paper that you like.

Now followup on that paper(s) by:

1) Look at the keywords for that paper (usually under the abstract). And search for other papers on the same topic with same key words
2) Look at the bibliography for the paper and select the papers you need to read to get enough background on the topic, as well as papers that are similar.
3) Repeat steps 1 & 2 for the new papers you've found

Note, the algorithm above has potential to be infinite, that is where I come in... Start with a topic, find a few papers, and then lets meet and I'll give you pointers on what to do next. Shortly you'll probably have to either refined or expanded the area of research, but you probably won't know until you find at least a few papers on something you are interested in.

As you read each paper, try to write a paragraph summarizing the paper, what problem it solves, where it falls apart. This will serve as a good refresher after reading so many papers! I usually write it & staple it to the printed version of the paper.

Also, it probably won't hurt to search Google or some other engine for topics as well, but beware of what sources you find. For our purposes, we are looking to learn from juried sources, not opinion pieces.