surfacesyntacticud / tools

5 stars 0 forks source link

Impersonal diathesis #5

Open perrier54 opened 4 years ago

perrier54 commented 4 years ago

In the impersonal diathesis, the subject of the sentence is an expletive. The relation between the main verb and the expletive subject is noted expl:subj in UD and subj:expl in SUD. Even if these relations are used in other cases, the conversion between them is not a problem. The subject in the canonical diathesis becomes the direct object in the impersonal diathesis

Example: Canonical diathesis: deux personnes arrivent (two people are coming) Impersonal diathesis: il arrive deux personnes (there are two people coming)

In SUD, the relation between the verb and the object, which is the subject in the canonical diathesis, is noted comp:obj@agent. In UD, it is noted csubj or nsubj as in the canonical diathesis, because the relation which is considered is deep subject.

When converting from UD to SUD, we must distinguish deep subjects in the impersonal diathesis from canonical subjects. In most cases, we can use the context to do this, but in some cases the expletive subject is far from the main verb, as in the following example.

Il peut être arrivé un malheur (Something bad may have happened) The expletive subject "Il" is distant from the verb "arrivé".

To facilitate the conversion, I propose to add extension impers to nsubj or csubj for the deep subjects in the impersonal diathesis. In the previous example, this would give: arrivé -[nsubj:impers]-> malheur.

The impersonal diathesis may be combined with the passive diathesis, as in the following example.

Il peut être demandé beaucoup d'attention (It may require a lot of attention)

The object of "demandé" is "beaucoup". The relation is currently noted comp:obj@pass in SUD and nsubj:pass in UD. I propose to replace nsubj:pass (csubj:pass) by nsubj:impers (csubj:impers) to distinguish this function in the impersonal diathesis of the canonical passive subject and to facilitate the conversion from UD to SUD. The passive voice can be recovered with the passive auxiliary.

sylvainkahane commented 4 years ago

I think it is interesting and useful to encode redistributions. It is done for passive and it could be done for impersonal. For passive, the marker is the auxiliary and we have the relations comp:aux@pass and comp:obl@agent. Sometimes the subject is also marked in UD by a relation nsubj:pass and we have a subj@pass in SUD, especially when there is no auxiliary for the passive.

According to this, I find ok to mark the subject in impersonal constructions by subj@impers. In impersonal passive (il a été demandé beaucoup d'attention), we will use both subj@impers and comp:aux@pass.

il arrive deux personnes : arrive -[subj@impers]-> il

We have problem with impersonal passive

perrier54 commented 4 years ago

The comment from @sylvainkahane doesn't exactly answer the question I asked. In the conversion between UD and SUD, there is no problem concerning the impersonal subject. In SUD, the relation is noted subj@expl and in UD expl:subj. The label subj@impers proposed by @sylvainkahane is subsumed by subj@expl, which is more general. Why not ?

What is problematic is the direct object, which represents the subject in the canonical diathesis. The conversion from UD to SUD is complicated by the fact that UD uses the same relation name for the subject in the canonical diathesis and the object in the impersonal diathesis, hence my proposal.

For simple impersonal redistribution, the subject becomes the direct object but it is always noted nsubj or csubj in UD. For conversion to SUD, we must distinguish it from a canonical subject. I therefore propose to add the extension impers. In SUD, the relation is currently noted as comp:obj@agent. If we decide to mark the impersonal redistribution in all cases, we can note the relation as comp:obj@impers .

Example: il arrive deux personnes (there are two people coming) In UD: arrive -[nsubj:impers]-> personnes arrive -[expl:subj]–> il In SUD: arrive -[comp:obj@impers]-> personnes arrive -[subj@expl]–> il

For passive impersonal redistribution, the object remains the direct object but it is noted nsubj:pass or csubj:pass in UD. For conversion to SUD, we must distinguish it from a canonical subject. I therefore propose to add the extension impers instead of pass. There remains a mark of passive redistribution in the relation between the verb and the auxiliary. If we decide to mark the impersonal redistribution in all cases in SUD, I propose to note the relation as comp:obj@impers in SUD.

Example: il a été vendu deux voitures aujourd'hui (two cars were sold today) In UD: vendu -[nsubj:impers]-> voitures vendu -[expl:subj]–> il In SUD: vendu -[comp:obj@impers]-> voitures a -[subj@expl]–> il

For passive pronominal impersonal redistribution, I propose to ignore the passive pronominal redistribution and to take only the impersonal distribution into account. The passive will be marked with the reflexive pronoun.

Example: deux voitures se sont vendues aujourd'hui (two cars were sold today) In UD: vendues -[nsubj]-> voitures vendues -[obj:pass]-> se In SUD: vendu -[subj]-> voitures vendues -[comp@pass]-> se

Example: il s'est vendu deux voitures aujourd'hui (two cars were sold today) In UD: vendu -[nsubj:impers]-> voitures vendu -[expl:subj]–> il vendu -[obj:pass]-> s' In SUD: vendu -[comp:obj@impers]-> voitures est -[subj@expl]–> il vendu -[comp@pass]-> s'

perrier54 commented 4 years ago

I go back to my last proposal. UD marks passive redistribution. In impersonal passive redistribution, it is not possible to mark both redistributions. In order to comply with the UD guidelines, I propose to dispense with marking the impersonal redistribution.Thus, the previous annotation examples become:

Example: il arrive deux personnes (there are two people coming) In UD: arrive -[nsubj]-> personnes arrive -[expl:subj]–> il In SUD: arrive -[comp:obj@agent]-> personnes arrive -[subj@expl]–> il

Example: il a été vendu deux voitures aujourd'hui (two cars were sold today) In UD: vendu -[nsubj:pass]-> voitures vendu -[expl:subj]–> il In SUD: vendu -[comp:obj@pass]-> voitures a -[subj@expl]–> il

Example: il s'est vendu deux voitures aujourd'hui (two cars were sold today) In UD: vendu -[nsubj]-> voitures vendu -[expl:subj]–> il vendu -[obj:pass]-> s' In SUD: vendu -[comp:obj]-> voitures est -[subj@expl]–> il vendu -[comp@pass]-> s'

This annotation complicates the conversion from UD to SUD in some cases. See the following example.

Example: Il va pouvoir être vendu deux voitures aujourd'hui (It's gonna sell two cars today) In UD: vendu -[nsubj:pass]-> voitures va -[expl:subj]–> il In SUD: vendu -[comp:obj@pass]-> voitures va -[subj@expl]–> il

Example: Deux voitures vont pouvoir être vendues aujourd'hui (two cars can be sold today) In UD: vont -[nsubj:pass]-> voitures In SUD: vont -[subj]-> voitures

The same UD relation nsubj:passis converted to comp:obj@pass or subj depending on the context, which is not possible to express in a single conversion rule.