generic infinitive verbal forms

[sylvainkahane]

In French or English, most infinitive verbal share their logical subject with another argument in the clause (raising or equi-construction), which is marked in SUD with the deep feature @x. In some cases, the infinitive verbal form does not and it has a generic interpretation:

il faut travailler plus
fumer n'est pas recommandé

I think that these rare cases should be marked rather than the default case. I propose to introduce a feature on generic verbal forms. It will be redundant with the absence of @x, but it will be more reliable.

[perrier54]

As it is redundant with the absence of @x, I don't understand the point of adding this new feature. It would be interesting if there were clearly delineated cases where infinitives do not have an explicit logical subject and are not generic.

[sylvainkahane]

Redundancy is not a problem. In our annotation of idioms with features both on relation and nodes, we opted for a redundant annotation.

In this case the redundancy is less evident. For instance, if you want to know the ratio of ininitive verb with a generic interpretation it is not trivial (if someone can write the good request to get the generic infinitive verbs, I'm interested).

I bet that less than 10% of infinitive verbs are generic. And it is for this reason that I think it is much more efficient to annotate them, rather than annotating the ones that have a logical subject. I think they will be less errors in the annotation if we do that.

[perrier54]

I understand @sylvainkahane better. The only objection is that this could make the conversion between UD and SUD more complicated because in UD, xcomp is used for verbs whose subject is controlled by another verb.

Remark: finding the ratio of infinitive verbs with a generic interpretation is not very difficult with the current annotation, but it can depend on the language. For SUD_French-GSD, we need three GREW requests to find all verbs in the infinitive with their controlled subject :

pattern { GOV -[deep=x]-> DEP; DEP[VerbForm=Inf] } pattern { GOV -[deep=x]-> DEP; DEP1 -[comp:obj]-> DEP; DEP[VerbForm=Inf] } pattern { GOV -[deep=x]-> DEP; DEP1 -[comp:obj]-> DEP2; DEP; DEP1 -[comp:obj]-> DEP2; DEP[VerbForm=Inf] }

[Julie921]

Obsolete annotation. See surfacesyntacticud/guidelines#29