#222: More Citation metrics
-------------------------+--------------------------------------------------
 Reporter:  tbrooks      |       Owner:     
     Type:  enhancement  |      Status:  new
 Priority:  major        |   Milestone:     
Component:  WebSearch    |     Version:     
 Keywords:  INSPIRE      |  
-------------------------+--------------------------------------------------
 Taking advantage of author disambiguation, once it is in place, as well as
 the "refersto:" operators, we can imagine a number of new citation
 metrics:

 1) without self cites.   This was an option in bibrank, however, we should
 be able to do this by hand in the summarizer, since, in general one may
 want two different things:
 a) exclude cites to paper a by any author of paper a, then repeat for all
 papers in a set
 b) define a set of papers (say those by someone from SLAC) and find all
 cites not from that set.
 A and b are different in general, and b can only be done on the fly.

 2) source/sink stuff as on the alt/ cite metrics page (i.e. cites-refs,
 cites+refs)

 3) Number of citing papers for a given search. I.e. define a set of papers
 and count how many papers cite that set.   This is different than counting
 the number of papers that cite each paper within the set, and then adding,
 since papers may cite more than one paper in the set.   This metric would
 be interesting in terms of say how many papers an author has influenced,
 and would remove heavy weighting to people who are cited in large blocks
 by only a few papers.

 4) extending the above #3, one could count # of distinct authors citing a
 given set of papers, and also the distribution of those (i.e. this set of
 papers is cited 20 times by ellis, 19 times by peskin, and 1 time by 20
 other authors....)

 5) Keep thinking, the sky's the limit.   Not all of these are immediately
 suitable for putting out on the public web, but we should be toying with
 them...

-- 
Ticket URL: <http://invenio-software.org/ticket/222>
Invenio <http://invenio-software.org>

Reply via email to