@x vs Generic/Raising

[sylvainkahane]

@x has been introduced to be able to get back the UD distinction between xcomp and ccomp. Nevertheless for an infinitive verb in a dependent position, the default case is to be @x. It would be better to encode the particular case where there is no raising and the logical subject of the infinitive has a generic interpretation. I suggest to replace the @x feature on the relation by a feature on the verb itself. I don't know how to name this feature. It could be Raising=Yes|No or Generic=Yes|No or Subject=Raising|Generic. It could even possible to be more precise and to precise if it is SubjRaising, ObjRaising, or OblRaising:

Subject=SubjRaising: Peter wants to leave
Subject=ObjRaising: Peter wants Mary to leave
Subject=OblRaising: Pierre propose à Marie de venir
Object=SubjRaising, Subject=Generic: the book is tough to read
Subject=Generic: it is impossible to leave

If we only indicate Raising vs Generic, it can be done by a Grew rule.

[perrier54]

I dont' understand the advantage of replacing @x attached to the dependency from the main verb to the complement verb giving rise to raising with a feature attached to the main verb, because it is equivalent.

On the other hand, I see the disadvantage because @xis also used with mod. For example: Melanie will call me to confirm your lie.. call -[mod@x]-> to

In this case, it makes no sense to attach a feature to the verb call.

[sylvainkahane]

In your example, the feature would be added on the infinitive verb confirm. There are several advantages: 1) in SUD the @x feature is on the relation call -[mod]-> to, while it is above all a property of the infinitive confirm, whose logical subject is me 2) the default case for an infinitive is the raising case. For this reason it is better to indicate when the subject is generic, if we consider the generic vs raising property to be important. I suppose that @x has been forgotten in some examples. 3) if we note the raising it would be interesting to note if it is a subject or object raising (cf. Pierre promet à Marie de venir vs Pierre permet à Marie de venir) 4) we can extend the annotation to the object raising: the book is tough to read

[perrier54]

I seem to have misunderstood @sylvainkahane 's proposal. I understood that the Subject feature is attached to the governor verb and not to the infinitive. I agree with the proposal to attach the feature to the infinitive and prefer the more precise version, as it can be generalized to object raising like in the example the book is tough to read. We can attach the feature Object= SubjRaising to read, assuming that the adjective does have book as its subject. In this other example He is unable to work, we can attach the feature Subject= SubjRaising to work.

[bguil]

SUD French data have been updated but, when there is no overt raised subject:

• ParisStories and Rhapsodie use Subject=NoRaising
• GSD uses Subject=Generic

It is needed to decide one solution and to make corpora consistent.

[perrier54]

As NoRaising = Generic | Instantiated, the annotations are consistent. The annotation of French-GSD is more precise.