(Originally appeared in Lognet 93/2)

Progress Report on the Case-Tag Project

by Kirk Sattley

It is certainly long past time to issue a progress report on the case-tag project. James Jennings and Steve Rice have each sent long messages concerning general approaches and particular rules for deciding how to assign tags. I’ve been considering these, and working with the list of primitive predicates, to see if we’re reaching any sort of consensus on their taggings. Presumably, with a thousand samples to work from, the Dictionary Helpers will find their job easier. (Not forgetting that there are some questions about compounds that don’t arise with primitives.) I’ve also been trying to classify the problems I’ve encountered in thinking about tags, so as to write draft essays on what I hope are separable facets of the problem; I plan to send these out one at a time...soon.

James Jennings had the brilliant idea of extracting the tag patterns from RAM’s [Dr. Robert A. McIvor, our Chief Grammarian’s] file of primitives into a separate field, then sorting the list on that field to bring together the Prims with similar patterns of tags. Now (since there are more than a few days on which I find it easier to do manipulative busy work than serious thinking), I’ve taken this file, updated the definitions that have changed since NB3 (Notebook 3: The Present State of the Loglan Language, 1987), added the new primitives listed in Lognet’s “Lo Cninu Purda” up through LN 91/2, and set up three tag-pattern fields, for the NB3 taggings, Jennings’s, and my own. The records in the file thus look like this:

Prim NB3 JJ KS Definition

badjo B-F B-F B-F 1 is a bough/limb/branch

of 2 ...

bidje B-F* B-FN B-FN 1 is an edge of 2

between surfaces 3

(plural arg) [L1]

gramo B-* B- B-C 1 is/weighs 2 gram(s)

(default =1)

[L1: two args]

detra P-* ?-? P-S 1 is a daughter of 2 [L1: 2 args only]

[JJ:P-S/S-D]

garti D-SP K-DB K-DV 1 is grateful to 2 for 3

[JCB: S-DV]

[SLR: K-BS]

lansa *-* *-* P-S 1 is a spear/lance made by 2 [LN91/2]

nemdi D-KB ?-?? K-DV 1 is an/the enemy/foe of

2 in struggle 3

[JJ: D-SV/K-KB]

stuka F-B F-B B-S 1 is a stick of/made of 2

Whenever anyone else has suggested a tagging for a particular primitive,I’ve added it in bracketed notes. See, for instance, garti’. (JCB = James Cooke Brown; SLR = Stephen L. Rice.)

My own taggings, where they disagree with anyone else’s, are serious, but tentative of course. Every time I look at the list, I find something to change ... or to change back. I suggest this is a good time to revive the E-Mail dialogue.

I’d be happy to send this file, or a subset of it, to anyone who asks. The potentially interesting subsets, I think, are these:

(A) The whole file. This contains 1,018 primitives, is 1,107 lines in length in print format, and contains 52,078 bytes.

(B) The primitives which have been added, or whose definitions have changed, since NB3; these are the entries that have an asterisk in the NB3 field. This file has 106 primitives, 142 lines, and is only 6,319 chararacters in length.

(C) The primitives on which either Jennings or I disagree with the NB3 tagging (this includes (B) above, since at least I disagree with all asterisks). This has 304 primitives, 450 liines, 21,392 bytes.

There are several interesting ways to sort the file, as well:

(1) Alphabetically by primitive predicate.

(2) By tag pattern (alphabetically by the tag-code letter-string):

(a) NB3 tagging

(b) JJ tagging

(3) By number of arguments first, then sorted within each class by any of the preceding criteria.

Also, it is easy to structure the file in either of two formats:

(P) Print format (as in samples above): 80-character lines, tab stops every 8 characters; continuation lines of the Definition field headed with 4 tabs.

(R) “Record” format (for easier further processing, perhaps): tab-separated fields, definition field not broken into additional lines (longest record is about 120 bytes). The sizes above are for the bodies of the files, uncompressed. Each file also has a 965-byte prefix explaining the format and labelling the columns. For those on CompuServe who prefer binary E-mail, I can compress the file with ARC, ZIP, LHArc, or StuffIt.

Ae hapci penso,

Hue Krk