Input Inference

Looking

Searching for related work is non as trivial as it sounds, so I list down here few tips on how to deal with it.

• Same author, same topic. Authors tend to focus on a limited number of research topic, so if an author published a paper on a topic, it is likely that he/she published other papers as well. Similar approaches, better evaluations, extensions, and so on.
• Same conference (track), similar topic. Conferences are organized by topic and domain, so if a paper appeared in a conference, then other papers in the same conference (or in conferences on the same domain) likely published related papers.
• Citing and Cited by. Follow the conceptual links between the papers. What are the papers that are cited by the one you are reading (looking at the present/past), and what papers cite it (looking at the future).
• Keywords. Build up your own set of keywords that clearly focus on the specific topic you are looking for. You can start from the keywords listed at the beginning of the paper, but you definitively want to create a more refined set. Once you have a robust, yet concise, set of keywords, your queries will be more effective (better papers in less time)

Handling

Looking for related work is a process. You start by reading a paper, making yourself familiar with the background concepts and the context elements. Then you search more papers (see section above), read them, repeat. Eventually, you'll fill confident about what's already out there.

A possible way to proceed is to collect papers in a first session (just look at the time and/or abstract); then take a break, go through the elements which you have collected (read basically the intro) and decide if the paper is actually on point or not. Take a note about the reason why you discard the papers. As you go build a white-list, a gray-list, and a black-list of papers  (with annotations). Take a break, now is the time to actually read the papers (in the white-list first), make a short summary about the most interesting points and forget the implementation details for the moment. Update your lists, find more papers and repeat.

When to stop?

This is tricky... Eventually, you'll see that you reached a fixed point: all the papers that you find are already in one of your lists. This is a good indicator that you might stop, but do not rely on it all the time. Remember that you might have missed some important keyword or some domain, so do not give for granted that you know everything. Likewise, do not aim at knowing everything, when you feel confident "Please Staph!".

Use the right tool for the job

Start now using a Reference Manager system. There's plenty of them and most are free: Zotero, Mendeley, BibRef, JabRef, etc.. Many of those system have cool features: you can add notes about papers, tag them, connect one to another, and so on. Do not underestimate the possibility to automatically generate references (i.e., .bib files) out of it. Additionally, some of them let you share your library or meta-data out of it. This might come handy if you do group work for example.