Sometimes I need to quantify something that has a named predicate as its key relation. I don’t have any way to tell from that alone whether it’s the kind of proper noun that should take proper_q (like a name) or def_q (like a sports team).
I was hoping to just generate all possible options using def_or_proper_q which I found in the SEMI and let the decision of which output is best to be a problem for later.
But when I tried to generate with it I got this error:
NOTE: EP 'def_or_proper_q' is unknown in the semantic index
The use of abstract predicates in the SEMI is not so well documented, unfortunately. There are three obstacles to underspecifying the quantifier in the way you wanted: (1) the type def_or_proper_q, in spite of its tempting-sounding name, is not actually a supertype of the predicate for “the”, but instead has a more limited set of definite-quantifier subtypes, once used to deal with demonstratives combining with proper names as in “this Kim”, but in fact probably no longer used in parsing; (2) the grammar does not define the quantifier type hierarchy in such a way as to provide a supertype that subsumes both “_the_q” and “proper_q”; and (3) if there were such a type, it would need to be included with its subtypes in the file etc/erg.smi, where abstraction types deemed to be useful for generation are defined, as for example with the type def_udef_a_q as an abstraction subsuming the predicates for “the”, “a/an”, and the null determiner used for bare plurals and mass nouns.
In short, there is no straightforward way to underspecify the input to the generator to get both “the Browns” and “Browns” as well as “Kim” and “the Kim”. You might get close to what you want by exploiting the number property on the named entity, supplying the quantifier predicate num_or_proper_q if it’s singular, and def_explicit_q if it’s plural.