[matrix] need help on some implementation issues

[Altaf Mahmud]

Dear Madam:

Thanks for your reply. And again I have faced some problems and need help:

1. In LAB-06, 'adjective-relation' has been introduced for
LKEYS.ALTKEYSfeatures of demonstrative determiners. But, there is no
such type is
declared in 'matrix.tdl' file and LKB doesn't find it. What might goes wrong
here?

2. In LAB-06, in section for demonstrative pronoun, it says that the LBL
values of two relations (LKEYS.KEYREL noun-relation and
LKEYS.ALTKEYRELevent-relation) are to be identified with each other. I
think I did the same
thing as the following code (reentrancy with '#lbl' ):

;;;;;;;;;;;;;;;; pronoun lexeme type ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
pronoun-lex := sg-nouns &
[ SYNSEM [ LOCAL.CAT.VAL.SPR < [ OPT + ] >,
           LKEYS.KEYREL noun-relation &
                        [ PRED "_pronoun_n_rel" ] ] ].



;;;;;;;;;;;;;;; demonstrative pronoun ;;;;;;;;;;;;;;;;;;;
demo-pronoun-lex := pronoun-lex &
[ SYNSEM [ LOCAL.CONT.HOOK.INDEX.COG-ST activ+fam,
           LKEYS [ KEYREL.LBL #lbl,
                   ALTKEYREL event-relation &
                             [ LBL #lbl ] ] ] ].



;;;;;;;;;;;;;;;;; lexical entry ;;;;;;;;;;;;;;;;;
ei_demo := demo-pronoun-lex &
[ STEM < "ei" >,
  SYNSEM [ LOCAL.CONT.HOOK.INDEX.PNG [ PER third,
                                       GRD non-hon,
                                       TPC low ],
           LKEYS.ALTKEYREL.PRED proximal+dem_q_rel ] ].

It can parse, but it can't generate because it couldn't show any MRS for the
main transitive verb of the sentence. For example:

"naafid bhaat khaay"
"Naafid rice eats"
(Naafid eats rice)

There is no MRS for the main verb 'khaay' which should be 'eat_v_rel' and
handle 'h1', but the LTOP of the whole MRS is 'h1' as usual. Did I do any
mistake here about following the lab guidelines?

3. I 've written some morphological rules for verbs. Free word order
language 'Bangla' has subject-verb agreements only with person variations.
However, when I try to generate from a given parse tree, it only generate
the sentences with the right most verb form of the rule. An example
for 3rd-person
verb form at 'irules.tdl' file:

3p-verb :=
%suffix (* e) (* chhe) (* chhila) (* chhilo) (* echhe) (* echhila) (* be) (a
ay) (a acchhe) (aa eyechhe) (jaa giechhe) (aa eyechhila) (aa eyechhilo) (jaa
giechhila) (jaa giechhilo) (e ey) (e icchhe) (e icchhila) (e icchhilo) (e
iechhe) (e iechhila) (e iechhilo) (e ibe)
3p-verb-lex-rule.

Here, the right most 3rd person verb form is "(aa eyechhilo)" for those
verbs which have ending letters 'aa'. I know LKB checks the rules from right
to left, but it generates the sentences with only this verb form, not with
other verb forms at left. For example an input sentence: "naafid bhaat
khaay" (Naafid eats rice), it generates only:

"naafid bhaat kheyechhilo" (Naafid ate rice)
"naafid kheyechhilo bhaat"
............................

But, not any sentence like "naafid bhaat khaacchhe" (Naafid is eating rice).
Here, root form of the verb 'khaay' is 'khaa' at 'lexicon.tdl' file.


4. The last question here about single word parsing, of course I should ask
you long before, but I didn't even try with it. I can't parse a single word
yet, although I 've changed the 'globals.lsp' file as following:

(defparameter *start-symbol* '(root lex-root)
  "specifing valid parses")

No, I don't get the single word parse.


Thanks a lot for being patience to read such a long mail of me.


-- 
-Altaf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/matrix/attachments/20070715/08affdce/attachment.html>