[developers] ersatz+CARG

Ben Waldron bmw20 at cl.cam.ac.uk
Thu Oct 19 15:25:51 CEST 2006


It's about time I announced to the list the magic runes necessary to 
instantiate the CARG of an "preprocessor ersatz" with the surface string 
matched. Sometime in the future this "surface string" ought to be 
replaced with some kind of "normalised form" (eg. the same date may be 
written in multiple different ways, but ought to contribute the same 
semantics in each case). Please see below.

- Ben

>
> In the LKB, set the following:
>
> ;; insert surface form of ersatzes at end of this path
> (defparameter smaf::*ersatz-carg-path* '(SYNSEM LKEYS KEYREL CARG))
>
> If you alter this parameter at runtime it is necessary to call 
> (smaf::reset-conf) before the new setting will take effect.
>
> In PET, set the following:
>
> ;; insert surface form of ersatzes at end of this path
> ersatz-carg-path := "SYNSEM.LKEYS.KEYREL.CARG".
>
> The LKB treats anything created by an ersatz preprocessor rule (in 
> FSPP) as an ersatz; PET simply checks the last 6 characters of  an 
> input token for the suffix string "ersatz" (this will be fixed 
> properly later when the PET/SMAFh XML interface is complete).
>
> All ersatz entries in the lexicon should have unset CARG values. If 
> not, the LKB will spit out a warning, and reset the CARG to the 
> surface string; PET will spit out a warning but keep the _original_ 
> setting (this is probably the better behaviour).




More information about the developers mailing list