By the time the post processor gets its hands on a meaning form, the form may contain
ten kinds of literals (as documented in
deplambda.others.PredicateKeys
;
see also the patterns in PostProcessLogicalForm
. For a description of the types see
ud_types.txt.
We use the following types below:
tri
is<v,<v,<v,t>>>
b
is<v,<v,t>>
u
is<v,t>
v
is<a,e>
Below, arguments in {...}
are specified at run-time. I also try to specify information about some of
the meta-variables in the predicate description gleaned from examining the current (ud-
) rules
in lib_data. (Note:
the earlier deplambda-
rules in the same directory have a different syntax and correspond to an
older version of the system.)
p_EVENT.ENTITY_<Relation>:b
. In the current set of rules,<Relation>
is one ofl-nmod.{casemarker}
orl-nmod
arg1.{casemarker}
orarg0
,arg1
,arg2
.vp_EVENT.EVENT_{relation}:b
p_COUNT:b
p_TARGET:u
p_EVENT_{word}:u
.{word}
is typically of the formw-<Index>-<Word>
, where<Index>
is the integer index into the utterance, and<Word>
is the corresponding word.p_TYPE_{word}:u
p_TYPEMOD_{word}:u
p_EVENTMOD_{word}:u
p_CONJ:tri
p_EQUAL
(Note: This exists inPredicateKeys
but there are no instances of this in theud-
rules.)
Actually, the rules also produce
p_EMPTY:u
-- this is not documented inPostProcessLogicalForm