crs%lambda@LANL.ARPA (Charlie Sorsby) (12/10/86)
In article <8612090611.AA11211@ucbvax.Berkeley.EDU>, HAMILTON%RCSMPA@gmr.com ("William E. Hamilton, Jr.") writes: > I emphatically agree with the following comment from Robert Amsler: > > >Can something be done to minimize the idiosyncratic special character > >usage in the bibliographies? First, I would like to mention that I appreciate the fact that those who are kind enough to share their bibliographies with us may not be eager go to a lot of additional trouble to put them into some standard form. Yet, there is a problem that should be addressed if the posted bibliographies are to provide the greatest possible benefit to net readers. One thing that I would find helpful is a note (perhaps a single line) at the beginning of the article specifying the program for which the bibliography was formatted. Realize that a "common" system, that you may recognize instantly, may be totally foreign to someone else. A few simple lines such as: This bibliography is formatted for the "refer" program (See Unix User's Manual Reference Guide and Unix User's Manual Supplementary Documents) may make the article immensely more useful to someone. If the bibliography is written for a particular formatting program and macro package, say so in a similar line, again pointing the reader to a source of additional information, if possible. Remember that we haven't all seen every possible text formatter and bibliographic program. For moderated news groups, how about collecting bibliographic/formatting programs (where possible) or pointers to where they can be obtained (where not) and archiving them in some way that makes them accessible to the readership of the group. Perhaps this information should be archived in, say, mod.sources instead, with only a pointer in the group that archives the bibliography. Perhaps a "new-user's" article can be routinely, or even automatically, posted, say once a month, to tell new users about this information and where it can be found. Finally, how about some of you out there, who may have the time and expertise, writing some translators that will allow bibliographies in one format to be translated to another. Then, if one collects a bibliography in a format that doesn't match the program available, it can be translated. These filters can also be archived in mod.sources with pointers in the various groups to which bibliographies are posted. -- Charlie Sorsby crs%lambda@lanl.arpa crs%lambda@lanl.uucp [I'm not sure I could legally distribute the bib/refer sources, even if they were of use to anyone who didn't have Unix. I'll leave distribution to AT&T, mod.sources, or the program author. I have answered a few AIList-Request queries about the format, but mostly I have forwarded queries to Lawrence Leff, the compiler of most of the bibliographies. The topic codes, MAG abbreviations, etc., are all under his control. Lawrence has agreed that the tbl formatting in the last installment was not successful. I'm sure he will consider other suggestions for making the material more useful. Personally, I find it amazing that the bibliography format problem still exists. I propose that it should be considered one of the great unsolved problems of AI. Computational AI has sprung from a clique of hackers who develop symbol-processing hardware and languages for the sole purpose of making it easier to get their own work done (developing symbol-processing ...). They and others in CS have applied incredible talent to the parsing of programs, natural language, and other strings. They have given us a few good spelling checkers and perhaps three reasonably good text formatters, and now we are even getting elegant fonts and professional text layout. All this, and yet I have never heard of a program for parsing citations. The best we can do is to use macro expansions for different journals after we have keyed each field by hand. People can generally parse any citation without even knowing the order of the fields or which other syntactic conventions are used. Boldface helps, but even if it were not perceived by the scanner the computer should still be able to figure out volume number, pages, etc. Yet I doubt that this problem has been solved even for the case of citations in a known format. Once someone develops the algorithm (including rules for Col. T.-W. Alphonse de Leon III, Ph.D., with or without misspellings) we can all communicate our citations in human-readable form and let the machines decode them for database use. -- KIL]