[developers] pet/lkb inconsistency

[Francis Bond]

G'day,

we are having a wierd issue with cheap (0.99.14svn).

When we call the KRG (development branch, very much based on the
Matrix), the top predicate becomes type u, not e.  As shown below:


 [ LTOP: h1
   INDEX: u2 [ u E.ASPECT: ASPECT E.MOOD: DECL E.PASS: BOOL E.TENSE:
PRESENT SENSE.ADVERSITY: LUK SORT: SEMSORT SF: PROP-OR-QUES ]
   RELS: <
          [ "_개_n_rel"
            LBL: h3
            ARG0: x4 [ x PNG.NUM: NUMBER PNG.PER: NON-HUMAN SORT:
SEMSORT SPECI: BOOL COG-ST: COG-ST SENSE.ADVERSITY: LUK ] ]
          [ "exist_q_rel"
            LBL: h5
            ARG0: x4
            RSTR: h6
            BODY: h7 ]
          [ "_짖_v_rel"
            LBL: h8
            ARG0: u2
            ARG1: x4 ] >
   HCONS: < h6 qeq h3 > ]

The parse with the LKB is identical, except that u2 is e2 (as we would
expect, verbs are constrained to have their ARG0 as event).

This does not happen with Jacy.   Does anyone have any idea where the
problem may lie?  Grammar available from Sanghoun on request :-).

-- 
Francis Bond <http://www2.nict.go.jp/x/x161/en/member/bond/>
NICT Language Infrastructure Group