finin@antares.PRC.Unisys.COM (Tim Finin) (05/20/88)
We have a need to process bibliographic references, extracting the
relevant information encoded in them. That is, to take a reference like:
J. W. Wallis and Edward H. Shortliffe. Customizing
explanations using causal knowledge. In Bruce G. Buchanan and
Edward H. Shortiffe, editors, Rule Based Expert Systems,
Addison-Wesley, Reading, MA, 1984.
and to produce a data structure something like:
((type bookChapter)
(author "J. W. Wallis and Edward H. Shortliffe")
(title "Customizing Explanations Using Causal Knowledge")
(book (title "Rule-Based Expert Systems")
(publisher "Addison-Wesley")
(editor "Bruce G. Buchanan and Edward H. Shortliffe")
(year "1984")
(address "Reading, MA")))
Put simply, we want to develop a system thast does what BibTeX does,
but in reverse. It should work for references to a variety of types
of documents (e.g. journal articles, books, technical reports,
theses, etc), and bibliographic styles. It should have clear
domain-independant knowledge (e.g. "Edward" is a given name, MA can
be an abbreviation for Massachusettes which is the name of a state,
1984 is a good value for a year of publication, etc.) and
domain-dependant knowledge (e.g. what IJCAI means, that BBN is a
company which has a technical reports series, etc). This would ease
its porting from one domain (e.g. AI) to another (e.g. fluid dynamics).
Such a system would probably be an interesting application drawing on
aspects of computational lingusitics (e.g. parsing, sub-language
theory, proper name recognition), and knowledge-based expert systems
(e.g. expectation-driven parsing, domain modeling). I'm interested in
getting pointers to any research on systems like this. I can't recall
hearing of any.
Tim
Tim Finin finin@prc.unisys.com
Paoli Research Center ..!{psuvax1,sdcrdcf,cbmvax,bpa}!burdvax!finin
Unisys Corporation 215-648-7446 (o)
PO Box 517, Paoli PA 19301 215-386-1749 (h)