I would like to have a tool which, when given a keyword(s) (such as a Google Scholar search query string), would map out a citation network (graph where nodes are papers, and edges are citations from one paper to the other) and/or collaboration network (nodes are papers or authors, and they are connected if they share authors/have co-published) for the papers which match that keyword (ie. a graph of just the papers returned for a Google Scholar or PubMed query).
It would also be helpful if instead of a keyword, I could give the tool one or several key publications in the area, and let it map out the neighbors which are at most X hops away.
I intend to use this to get an overview of important publication in a given field. Therefore it is important for the tool to be relatively fast (not take hours for a given query) and use up to date information.
I know people do research on this sort of stuff, for instance to see how information flows between subfields of a scientific discipline, or which fields are closely related. For instance, see this webpage: http://users.dimi.uniud.it/~massimo.franceschet/jbc/bibliometrics.html
However, often they publish results which paint a broad overview of all publications. Suppose I want to read up on "protein folding prediction". Punching that into Google Scholar, I will get hundreds of thousands of results. They are sorted according to relevance and citation number, but it can still be difficult to tell how the papers in the long list relate to each other; a graph can often convey information that a list cannot.
So the question is, does any tool exist that I can use for such a purpose?
There is a similar question: Generating a Citation Graph from a set of PDFs It doesn't seem a clear solution to Michael's problem - but its trivial to convert a Google Scholar (etc.) results list into a list of papers, so any program that solves his problem probably solves mine.
" but its trivial to convert a Google Scholar (etc.) results list into a list of papers": AFAIK, it's not; there is no API for google scholar and its detects robots.
PubMed has an API which IS robot friendly (see https://www.ncbi.nlm.nih.gov/books/NBK25500/#chapter1.ESearch). It would be nicer if this tool worked for all disciplines, but if its biology only, it would be good enough for me. I think one can have the user enter captchas for a Scholar scraper as well - doesn't Hazing's publish or perish use Scholar data?