• Aucun résultat trouvé

Describing Direct Speech in the Syntactic Reference Corpus of Medieval French (SRCMF)

N/A
N/A
Protected

Academic year: 2021

Partager "Describing Direct Speech in the Syntactic Reference Corpus of Medieval French (SRCMF)"

Copied!
1
0
0

Texte intégral

(1)

Describing Direct Speech

in the Syntactic Reference Corpus of Medieval French (SRCMF)

Nicolas Mazziotta, Universität Stuttgart, Institut für Linguistik/Romanistik

nicolas.mazziotta@ulg.ac.be

We present the way direct speech or thought representation (Marnette 2005) is handled in SRCMF. Our main issue is a methodological one: what should be encoded about this kind of enunciative construction in a purely syntactic corpus?

1

SRCMF Project and Direct Speech

Main Goal of SRCMF.Each of the three functional components of language (grammatical, semantic and enunciative, Hagège 1999) should be encoded separately in order to guarantee their mutual independence. Hence, in SRCMF,

Annotation is as genuinely syntactic as possible

• SRCMF enhances the Base de Français Médiéval (Guillot et al. 2007) and the Nouveau Corpus d’Amsterdam (Stein/Kunstmann 2007). . .

• with dependencial syntactic annotation layer; • SRCMF focuses on dependents of the verb.

Direct Speech Representation Issue.Dicendi verbs are func-tional tools that introduce direct speech. From an enunciative point of view, it is clear that utterances can be nested but. . .

. . . Lack of grammatical markers means grammatical de-pendencies cannot be assessed.

2

Utterance Hierarchy: Enunciative Perspective

From an enunciative perspective, dicendi verbs and direct speech are

nested utterances(Ducrot 1984), wheredeictics and vocativesare com-mon grammatical clues.

Et Tristran dit: “Sire, g’en vois [. . . ]” [“And T. said: ‘Lord, I am leaving’] (Béroul, v. 596)

Nevertheless, in SRCMF,

Principle 0 – The relationship between the utterances is

described from an enunciative point of view, rather than a syntactic one.

Moreover, quoted speech sequences can become long narratives, where deictics are not present any more. . . :

et li dit: “Sire vos me demandastes ore la senefiance de ceste aventure [. . . ] Quant li peres des cielx vit qu’il avoit en terre si grant durté [. . . ] il envoia son fil [. . . ]” [“and he said: ’Lord, you asked for the meaning of this event. . . When The Father saw that crualty was on the Earth, he sent his son’] (Queste, 168c, 38)

3

Utterance Hierarchy: Syntactic Perspective

Structures lack formal markers.There are no observable gram-matical tool linking the verb and direct speech in our first example:

Et Tristran dit: “Sire, g’en vois [. . . ]”

Hence, there are two equally valid ways to analyse it:

• Either Sire, g’en voisdependson dit (Iordanskaja/Mel’čuk 2009; Van Raemdonck 2004) [T. → dit

[Sire, g’en vois]]

• Or Et T. dit and Sire, g’en vois are two independentsentences (Tesnière 1965; Wilmet 2003) [T. dit]

|

[Sire, g’en vois]

The Valency Problem.Linguistic competence is not properly avail-able.

• risk that interference from the MF spoken by the annotators causes an incorrect assessment of the valency of verbs (respon-dre “answer”, crier “cry”, etc.);

• better not to state an erroneous analysis.

No Syntactic Dependency.Hence, we do not state syntactic depen-dency here, leaving this field open for further investigations.

Principle 1 – There is no dependency between direct

speech and preceding dicendi verbs.

4

Quotative Parenthesis

General Presentation.Dicendi verb can interrupt the syntactic structure of quoted speech and form a “quotative parenthesis”:

Sire, fet li preudons, por quoi le demandez vos? [“Lord”, says the good man, “why do you ask?”] (Queste, 166b, 40) Three analyses are possible:

• the parenthesis actually depend on the quoted speech (traditional approach, Grevisse/Goosse 1993);

• the parenthesis and the quoted speech are independent (Tesnière 1965) and position is not syntactically relevant;

• the quoted speech depends on the verb of the parenthesis (Van Raemdonck 2004).

For SRCMF, thespecific position is a grammatical mark:

Principle 2 – Quotative parenthesis is a dependency of

the verb of the quoted speech.

Pleonastic Cases.Quotative parenthesis can be pleonastical: *leur dist: “Seigneur, fit il, nous serons tenu pour malvais” [“He said to them: ‘Lord, said he, we’ll be taken to be evil’”] (Clari, XII, 12)

Following principles 1 and 2 solves the problem on a syntactic point of view.

Dealing with Uncertainty.Nevertheless, quotative parenthesis hap-pen to appear at the very end of the direct speech:

“Ce verroiz vos bien”, fet ele [“‘You will see it for sure’, she says” ] (Queste, 160a, 10)

This kind of construction leads to the same question as previously: is there a dependency? Second position of the verb could be seen as a grammatical mark, but dicendi verbs are possible at the beginning of a sentence (Moignet 1988).

Principle 3 – When direct speech does not surround the

quotative parenthesis, there is no dependency.

Références

Documents relatifs

This distinguishes the SRCMF project from the first major syntactic resource for medieval French: the corpus Modéliser le changement: les voies du français (MCVF) 4 contains,

Another strategy consists of not unpiling but rather perceiving an utterance including a pile as a Directed Acrylic Graph (DAG), that is to say a graph in which the

We think that our structure is better defined than the deep-syntactic tree and it can play a central role in the definition of lexical functions and paraphrasing rules We do not

The prosodic annotation alphabet INTSINT was originally based on the descriptions of the surface patterns of the intonation of twenty languages (Hirst &

The annotation was implemented at the several levels: the particular usage of PM (PM); its functions in this particular usage (Function PM); the commentaries for in- troducing

Though many corpora (e.g. Sketch Engine [1], Russian National Corpus [2], [3], and some others) offer examples of correctly written text containing given words, they are only able

Please cite this article as: Huet, S., Gravier, G., Sébillot, P., Morpho-syntactic post-processing of N-best lists for improved French automatic speech recognition, Computer Speech

In this sense we rely in our argumentation on reductio ad impossibile: if we assume a continuum approach and compare different speech reporting constructions of our two languages to