How do I pick good keywords for literature alerts?

F'x has made a very informative post here about staying up to date with current research.

My question concerns the "publication alert" part of it. To be clear: I am asking about setting up an online service, such NCBI PubMed, to send you a notification (for instance, email) whenever a paper matching certain search criteria is published.

The objective is to get a decent number of notifications, but not so many that you are swamped and unable to actually pay attention to them. Also, the notifications that you do get should be prioritized the most important papers.

The biggest problem I am having is coming up with effective keywords. If I try to make myself come up with general keywords (such as those that might be given in the article's own "keywords" section), it seems that I end up with keywords so general that I would get thousands of notifications every week.

Alternatively, I could try to come up with very specific keywords, but how do I select those? What if they are too specific, and an important paper gets published without using my particular obscure keyword?

Surely people are not born with a natural talent for thinking of good keywords. So how can I train this skill?

Furthermore, is there a systematic method to deduce good keywords given a body of "gold standard" relevant literature? (I mean things such as statistical analysis of the text)

I know that it can be worthwhile to focus on certain "big player" authors. But if in a field with thousands of authors (eg. cancer), how does a new researcher map out who is important and who is not? Furthermore, the biggest players are probably PIs of huge labs, who already publish a very large volume of papers, so we are again at odds with the "not too many notifications" problem.

Another possibility is looking at citation metrics to identify the most important papers, as Google Scholar does. But clearly, this is not effective for new papers - who will cite a paper published last week?

Answer

One of the biggest problems is that there are, in fact, waaay to many relevant papers coming out; that's because there are lots of articles published each month in many different fields.

The approach I found most useful was threefold:

• Set up author-specific alerts for the big names in your field. Read both those articles and the citations. They know the literature and will cite stuff you should be familiar with.


• Set up alerts to see when someone cites a seminal paper. In my research, there were certain seminal papers that everyone participating in my tiny subfield would always quote. Find those and set up alerts when someone cites that.


• Read the most popular article(s) of the month (most downloaded/shared/read in your journal of choice) to stay on top of interesting stuff that's going on.