[developers] Issues appending diff-lists

[Zhen Zhen Fan]

Dear developers,

We'd like to seek your help for this problem of dealing with diff-lists.

Chinese interrogatives have in-situ wh-questions, which have wh-words
appearing at positions of arguments, specifiers, and modifiers. While
implementing the analysis for these interrogatives, we

1. Define all wh-words to have its QUE containing an index (<! #index !>),
and all other words to have QUE as 0-dlist.

2. Redefine SYNSEM.NON-LOCAL.QUE to be a diff-list instead of 0-1-dlist (to
allow more than one wh-word to appear in a question), and make binary rules
to append QUE values of their daughters, as shown below:

basic-binary-phrase :+
  [ SYNSEM.NON-LOCAL.QUE [ LIST #first,
               LAST #last ],
    ARGS < sign & [ SYNSEM.NON-LOCAL.QUE [ LIST #first,
                           LAST #middle ] ],
       sign & [ SYNSEM.NON-LOCAL.QUE [ LIST #middle,
                           LAST #last ] ] > ].

3. Define the rule to identify clauses with at least one wh-word and set SF
to "ques". The problem is how to define the constraint for a diff-list with
at least one item inside.
We tried to use "QUE.LIST 1-plus-list", which works great to exclude
sentences with 0 wh-word, and to parse sentences with 2 or more wh-words.
For sentences with exactly 1 wh-word, some can be parsed and some can not.
It turns out that it works correctly if the wh-word is the first ARG in the
binary rule, and it won't parse if the wh-word is the second ARG in the
binary rule.
So it seems that appending gives us different results for the two scenarios:
a) when wh-word is the 1st ARG, mother's QUE can unify with the constraint.
-- 1st ARG's QUE: diff-list
                 [ LIST < 6 > + 16LIST,
                   LAST 16 ]
-- 2nd ARG's QUE: diff-list
                  [ LIST 19 0-1-list,
                    LAST 19 ]
-- mother's QUE: diff-list
                 [ LIST <4> + 18 0-1-list,
                   LAST 18 ]
b) when wh-word is the 2nd ARG, mother's QUE can't unify with the
constraint, complaining the conflict between 1-list and 1-plus-list.
-- 1st ARG's QUE: diff-list
                  [ LIST 28 0-1-list,
                    LAST 28 ]
-- 2nd ARG's QUE: diff-list
                 [ LIST < 6 > + 16LIST,
                   LAST 16 ]
-- mother's QUE: diff-list
                 [ LIST 1-list <4> + 11,
                   LAST 11 ]

How can this be resolved so that we get consistent results regardless of
the position of wh-word in a binary rule?

We also notice that two daughters with QUE as 0-dlist lead to the mother's
QUE as diff-list (not 0-dlist) with identical LIST and LAST.

We have also tried defining 1-list to inherit from 1-plus-list too. Then
the above problems will disappear, but it will fail to exclude sentences
containing no wh-word.

Many thanks!
Zhenzhen
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20161130/401b8094/attachment.html>