The ADS Bibliographic Reference Resolver

The ADS Reference Resolver is a service provided by the NASA Astrophysics Data System which allows users and information providers to query the ADS databases to verify the existance and availability of a particular bibliographic reference. The resolver attempts to be as flexible as possible in parsing reference strings and identifying the respective records, so that different citation formats are supported. Examples of references that are correctly parsed and identified include (but are not limited to):

• Journals in "modern" Astronomical style: Abt, H. 1990, ApJ, 357, 1
• Journals in "modern" Physics style: J. B. Gupta, K. Kumar, and J. H. Hamilton, Phys. Rev. C 16, 427 (1977).
• Journals in "Annual Reviews" style: Adainson M, Kerr Th, Whittet DCB, Duley WW. 1994. MNRAS 268:705-8
• Books (provided they are in ADS): Wali, K. 1991, Chandra: A Biography of S. Chandrasekhar (Chicago: Univ. Chicago Press)
• Conferences and conference series (provided they are in ADS): Rees, M. J. 1984, in Formation and Evolution of Galaxies and Large Scale Structure in the Universe, ed. J. Audouze and T. T. Van (Dordrecht: Reidel), 271
• Various small publications: Hermsen, W., et. al. 1992, IAU Circ. No. 5541
• Ph. D. theses: Pollock, J. T. 1982, Ph. D. Thesis, University of Florida.

The resolver processes the input strings by performing the following steps:

1. Reference parsing, i.e. identifying the different tokens listed in reference entry.
2. Reference identification, i.e. using the parsed tokens to generate the unique bibliographic code identifier for the reference.
3. Reference verification, i.e. querying the ADS database to verify the existence of the reference.

Often the resolver iterates on these steps in an attempt to match input records heuristically when an exact match is not possible or when "noise" is found in the input text.

We currently provide a form-based interface to the resolver as a prototype system simply for the purpose of allowing publishers to test its capabilities and offer suggestions or criticism. The output of the web interface consists of an unordered list (HTML <UL>) of list items that give a bibcode (or None, if the input line was totally illegible to the resolver), a confidence in parenthesis and then the original reference after a literal from.

Only bibcodes with positive confidence values should be used; the higher the confidence number, the lower the resolver's confidence in its result -- think of it as a "mangle number", describing how hard the resolver had to try to make sense of the reference. Note that books always have confidence 4, because we match books rather leniently (and have to, because of the relatively unstandardized way people cite books). If the reference could not be parsed (bibcode None), the confidence is zero, if the reference could not be verified (the resolver could come up with a bibcode, but could not find it in the ADS, or the article corresponding to the bibcode does not seem to be the one mentioned in the reference), the confidence is negative. In the latter case, the bibcode is more often than not completely bogus -- again, don't use them.

We also provide a reference PERL library and example scripts that can be used for automated batch resolution of references (although you'll probably need to modify the scripts to fit your needs). If you plan to use the script or a variation of it, please let us know so we can keep track of our user base in case we need to make changes to the interface. Here is some more information about the software:

• The README file
• An example script that uses the library, with typical input and output files.