Search

International audience; Conceived as a general introduction to syntax, this book presents the basic concepts necessary for a study of the combination of lexical and grammatical units within an utterance. The book does not impose a preconceived framework, but rather examines the various possibilities for the representation of syntactic structures, according to the general principles and specific criteria that have been adopted.The aim of this book is to provide a basis for teaching syntax at university, and to show that it is possible to identify the properties of languages in a methodical way and to put order in the jungle of each language. The book is divided into three parts: How to develop the model of a language? How to determine the basic units of a language according to their meaning, form, and combinatorial nature? How to define and represent the different ways in which the units are organized? This last part presents an abundance of syntactic diagrams of a wide range of types.The book is divided into small sections, alternating the main content with insights, historical notes, formal elaborations, linguistic examples in diverse languages, proposals for further reading, and exercises with answer keys.; Conçu comme une introduction générale à la syntaxe, cet ouvrage présente les notions de base nécessaires à une étude de la combinaison des unités lexicales et grammaticales au sein d’un énoncé. Sans se placer dans un cadre préconçu, l’ouvrage étudie les différentes possibilités pour la représentation des structures syntaxiques, en fonction des principes généraux et des critères particuliers retenus.Élaboré avec l’objectif de fournir une base pour l’enseignement de la syntaxe à l’université, cet ouvrage souhaite montrer qu’on peut dégager de manière méthodique les propriétés des langues et mettre de l’ordre dans la forêt vierge que constitue chaque langue. Il est divisé en trois parties : comment élaborer le modèle d’une langue, comment déterminer les unités de base de la langue en fonction de leur sens, forme et combinatoire, comment définir et représenter les différents modes d’organisation des unités. Cette dernière partie présente une abondance de diagrammes syntaxiques de diverses natures.L’ouvrage est découpé en de petites sections, alternant le contenu principal avec des éclairages, des notes historiques, des élaborations plus formelles, des exemples linguistiques dans diverses langues, des propositions de lectures additionnelles et des exercices avec des éléments de correction.

version modifiée le 14 avril 2023, correction des typo à la page 8, pas d’autres modifications, à part la mise en page; International audience; This paper introduces a typometric measure of flexibility, which quantifies the variability of head-dependent word order on the whole set of treebanks of a language or on specific constructions. The measure is based on the notion of head-initiality and we show that it can be computed for all of languages of the Universal Dependency treebank set, that it does not require ad-hoc thresholds to categorize languages or constructions, and that it can be applied with any granularity of constructions and languages. We compare our results with Bakker’s (1998) categorical flexibility index. Typometric flexibility is shown to be a good measure for characterizing the language distribution with respect to word order for a given construction, and for estimating whether a construction predicts the global word order behavior of a language.

International audience

National audience; This work presents a method and a tool for automatic extraction and exploration of statisticallysignificant patterns and potential grammar rules from treebanks.; Ce travail présente une méthode et un outil d’extraction et d’exploration automatique de motifs statistiquement significatifs, de potentielles règles de grammaire, à partir de corpus arborés.

International audience; The paper brings to the fore some advantages to first develop a new treebank in Surface-Syntactic Universal Dependencies (SUD) annotation scheme, even if the goal is to obtain a UD treebank. Theoretical benefits of SUD are presented, as well as UD-compatible SUD innovations. The twoway UD , SUD conversion is explained, as well as the possibility to customize the conversion for a given language. The paper concludes by a practical guide for the development of a SUD treebank.

International audience; The paper presents the first UD treebank for Beja, a Cushitic language spoken in Sudan. It has been built from the conversion and enhancement of an Interlinear Glossed Text (IGT). The paper's objectives are threefold: we explain our choice to use a morph-based annotation and its consequences, we describe the processing chain from an IGT to a morph-based dependency treebank and a word-based treebank, and we present several interesting constructions in Beja.

http://www.projet-rhapsodie.fr/; Rédacteurs : Sylvain Kahane (en collaboration avec l'ensemble de l'équipe syntaxe Rhapsodie et en particulier Kim Gerdes, Paola Pietrandrea et Christophe Benzitoun) Version révisée par Rachel Bawden et les autres annotatrices (Marie-Amélie Botalla et Adèle Désoyer)

https://www.projet-orfeo.fr/guide_annotation_octobre2017/GuideOrfeo-dependances_octobre2017.html

http://www.projet-rhapsodie.fr/; Rédacteurs : Sylvain Kahane (en collaboration avec l'ensemble de l'équipe syntaxe Rhapsodie et en particulier Kim Gerdes, Paola Pietrandrea et Christophe Benzitoun) Version révisée par Rachel Bawden et les autres annotatrices (Marie-Amélie Botalla et Adèle Désoyer)

International audience; In this paper, we approach the study of the verbal particles of Wolof by placing ourselves resolutely in a distributional approach based on the exhaustive study of a corpus annotated in dependency syntax by Dione (2019). This allows us to revise previous works by clearly distinguishing, among the verbal particles, a group of 6 particles (a, la, na, da, -u, ngi), which we will call assertive particles. The constructions in which these particles enter are characterized by a common distribution (impossibility to be in relative clauses and obligation to be in certain complement clauses) and by acommon topological structure characterized by the position of the assertive particle, the clitic complements and the imperfective auxiliary di. We propose a complete description of the combinativeness of all verbal particles, as well as the set of topological schemes of the predicative constructions of Wolof.; Dans cet article, nous abordons l'étude des particules verbales du wolof en nous plaçant résolument dans une approche distributionnelle basée sur l'étude exhaustive d'un corpus annoté en syntaxe de dépendance par Dione (2019). Ceci nous permet de réviser les travaux antérieurs en distinguant nettement, parmi les particules verbales, un groupe de 6 particules (a, la, na, da,-u, ngi), que nous appellerons les particules assertives. Les constructions dans lesquelles entrent ces particules se caractérisent par une distribution commune (impossibilité d'être dans les subordonnées relatives et obligation d'être dans certaines complétives) et par une structure topologique commune caractérisée par la position de la particule assertive, des clitiques compléments et de l'auxiliaire imperfectif di. Nous proposons une description complète de la combinatoire de l'ensemble des particules verbales, ainsi que l'ensemble des schémas topologiques des constructions prédicatives du wolof.

International audience; In this paper, we approach the study of the verbal particles of Wolof by placing ourselves resolutely in a distributional approach based on the exhaustive study of a corpus annotated in dependency syntax by Dione (2019). This allows us to revise previous works by clearly distinguishing, among the verbal particles, a group of 6 particles (a, la, na, da, -u, ngi), which we will call assertive particles. The constructions in which these particles enter are characterized by a common distribution (impossibility to be in relative clauses and obligation to be in certain complement clauses) and by acommon topological structure characterized by the position of the assertive particle, the clitic complements and the imperfective auxiliary di. We propose a complete description of the combinativeness of all verbal particles, as well as the set of topological schemes of the predicative constructions of Wolof.; Dans cet article, nous abordons l'étude des particules verbales du wolof en nous plaçant résolument dans une approche distributionnelle basée sur l'étude exhaustive d'un corpus annoté en syntaxe de dépendance par Dione (2019). Ceci nous permet de réviser les travaux antérieurs en distinguant nettement, parmi les particules verbales, un groupe de 6 particules (a, la, na, da,-u, ngi), que nous appellerons les particules assertives. Les constructions dans lesquelles entrent ces particules se caractérisent par une distribution commune (impossibilité d'être dans les subordonnées relatives et obligation d'être dans certaines complétives) et par une structure topologique commune caractérisée par la position de la particule assertive, des clitiques compléments et de l'auxiliaire imperfectif di. Nous proposons une description complète de la combinatoire de l'ensemble des particules verbales, ainsi que l'ensemble des schémas topologiques des constructions prédicatives du wolof.

International audience; This paper presents practical and theoretical guidelines for the development of treebanks for spoken languages in the UD and SUD annotation schemes. We discuss text-sound alignment, segmentation into "sentences", use of "punctuation", paradigmatic lists, disfluencies, and paratactic constructions. This proposal is based on the development of (Surface-Syntactic) Universal Dependencies treebanks for spoken French, Naija, and Beja.

International audience; This article considers the annotation of subjects in UD treebanks. The identification of the subjectposes a particular problem in Wolof, due to pronominal indices whose status as a pronoun or apronominal affix is uncertain. In the UD treebank available for Wolof (Dione, 2019), these have beenannotated depending on the construction either as true subjects, or as morphosyntactic features agreeingwith the verb. The study of this corpus of 40 000 words allows us to show that the problem is indeeddifficult to solve, especially since Wolof has a rich system of auxiliaries and several basic constructionswith different properties. Before addressing the case of Wolof, we will present the simpler, but partlycomparable, case of French, where subject clitics also tend to behave like affixes, and subjecthood canmove from the preverbal to the detached position. We will also make a several annotationrecommendations that would avoid overwriting information regarding subjecthood.

International audience

International audience; Why you should turn SUD-The importance of choosing a Surface-Syntactic dependency annotation scheme. The article attempts to promote the Surface-Syntactic Universal Dependencies (SUD) annotation scheme to syntactic annotation projects, as an alternative to the standard Universal Dependencies (UD) scheme, particularly on oral or non-standard texts, conducted for comparative and typological studies.; L'article défend le schéma d'annotation Surface-Syntactic Universal Dependencies (SUD) comme alternative au schéma standard des Universal Dependencies (UD) pour des projets d'annotation syntaxique, en particulier sur des textes oraux ou non-standard, menés dans un objectif comparatif et typologique.

International audience; This paper highlights the advantages of not interpreting connections in a dependency tree as combinations between words but of interpreting them more broadly as sets of combinations between catenae. One of the most important outcomes is the possibility of associating a connection structure to any set of combinations assuming some well-formedness properties and of providing a new way to define dependency trees and other kinds of dependency structures, which are not trees but "bubble graphs". The status of catenae of dependency trees as syntactic units is discussed.

International audience

International audience; This chapter is devoted to the presentation of the tools and methods used for the different steps of the semi-automatic syntactic annotation: automatic preprocessing; microsyntactic parsing with the FRMG tool, correction of the parsing with the Arborator tool, agreement analysis, post-validation correction, and development of the final format of the Rhapsodie syntactic treebank. As FRMG is a parser for written French that was not configured to analyze disfluencies and reformulation, we used our manual pile marking to unfold the piles and produce a series of simplified “sentences” with only government relations. Despite having two annotators plus a validator for the corrections, we found a substantial number of errors in the post-validation procedure by using a set of rules to determine the well-formedness of the trees.

International audience; This chapter describes the microsyntactic analysis of the Rhapsodie corpus in terms of dependency syntax. Microsyntax studies the relations between words that are characterized by a strong syntactic cohesion, traditionally called “government”. The different steps in the annotation are presented: segmentation into words and labeling in parts of speech, dependency structure, and basic syntactic functions. We justify our decision to use a small set of tags without redundancies, but to introduce a predicative relation for elements that form a complex predicate including copula, auxiliaries, and some modal verbs. Complex cases of annotations such as extraction (relative and interrogative clauses, cleft sentences) and negation are also presented. In addition, we show how a constituent structure can be computed from the dependency structure.

International audience; This chapter presents phenomena we call “piles” or “lists”, which are characterized by the fact that a list of elements piles up in the same syntactic position. We therefore group the analysis of coordination together with the analysis of other phenomena such as reformulation, disfluency, partial answer, or negotiation. The elements of a pile are linked to one another by a relation that is both syntagmatic (they follow one another) and paradigmatic (they fill the same syntactic slot with respect to their common governor). The syntactic analysis of the other elements – junctors, paradigmatic adverbs, and list completers – is discussed. We also propose a typology of the different cases of pile structure and introduce the seven subcases of paradigmatic links taken into account in the annotation.

International audience; In this paper, we show that building a treebank can be used as a way to establish a language. Annotated corpus can be used as tools when arguing that some linguistic data belongs to a separate language (rather than a dialect or variety of another established language). We provide here a case study on a treebank of Naija, a Post-creole spoken in Nigeria which presents us with significant differences from treebanks of English in terms of existing constructions and frequency of several syntactic units.

International audience; In this paper, we show that building a treebank can be used as a way to establish a language. Annotated corpus can be used as tools when arguing that some linguistic data belongs to a separate language (rather than a dialect or variety of another established language). We provide here a case study on a treebank of Naija, a Post-creole spoken in Nigeria which presents us with significant differences from treebanks of English in terms of existing constructions and frequency of several syntactic units.

International audience; This article relates to a study that deals with the linguistic analysis of proform in French Sign Language (LSF). To our knowledge, these structures remain poorly defined and are almost never taken into account in computer modelling, which is the longer-term goal of our studies that aims to design Sign Languages (SL) automatic processing. Given their rather frequent use in SL discourses, it seems necessary to us to study the different linguistic functions that they can take and the most frequent observable forms associated with these functions. This requires us to study which body articulators are involved (manual and non-manual) and how they are organised within spatiotemporal structures. In this article, we present the results of an exploratory study concerning one of these functions, the use of proforms to introduce an entity into the discourse. We have carried out a corpus analysis which revealed particular structures associated with this function and not described so far. We have modelled them using a formalism that takes into account the multilinear, spatiotemporal and iconic structure of SL, and which can be used in automatic processing.For us, a "proform" is a conventional handshape, generally motivated and iconic, representing an entity that has been already mentioned in the discourse or not. The proform takes its full meaning only in context. We use the term proform without a priori regarding the level of language (morphology or syntax) or linguistic functionalities. A number of linguistic functions have been mentioned in the literature, coming with a variation in the terminology (“classifier”, “polymorphemic predicate”, productive sign”, etc). We have tried to identify the different functions of the proforms mentioned in the literature, which could be:classifying function: An ambivalent function, since "classifying" can mean both the semantic class or the classes of forms, depending on the theory.adjectival function: These are mainly descriptors or size and/or shape specifiers.anaphoric function: In the case where the proform is used as an entity already mentioned in the discourse, often in tandem with a relevant spatialisation.spatio-locative function: Often used in the case of classifier constructions.presentative function: Presented as the case of nominal classifiers that introduce an entity into the discourse.In our study, our aim was to examine in detail this last function, which is how the proforms are performed when introducing a new entity in a discourse. We wanted to determine which forms correspond to this function. For this, we adopted a corpus-based methodology. The main corpus that has been used is the MOCAP1 corpus, a corpus of LSF created to conduct multidisciplinary studies. One of the tasks of this corpus was to describe 25 different photographs. These descriptions include many uses of proforms with different linguistic functions, including the one we are interested in. We then proceeded to annotate the corpus by identifying the temporal segments corresponding to the activity of the manual articulators (lexical signs, pointing, illustrative structures including the use proforms, etc) and non-manual ones (head, eyebrows, closing of the eyes, direction of gaze, cheeks, lips).For this study, we took into account the annotations carried out on 5 of the images for the 8 signers (40 videos) and carried out a first step of essentially qualitative analysis. These images were chosen for the greatest richness they generated in the descriptions. We do not yet have at this stage elements of a statistical nature to give. Statistical analysis would require extending the annotation to other videos in the corpus and other types of discourses in order to have enough occurrences for each observed phenomenon. At this stage, our aim is mainly to identify and describe structures that seem to be frequently produced by the signers.In order to describe the recurrent patterns observed, we used a formal representation called AZEE which takes into account the specificities of signed languages. AZEE was chosen among other systems because it allows to represent the activity of the different manual and non-manual articulators with the necessary and sufficient constraints for their realisation in any spatial and temporal context. This representation also aims to be implementable and thus usable for sign language processing.We focused particularly on three structures:First, the introduction of an entity with a “buoy” (perseveration of a handshape coming from a gestural unit previously signed). We observed that the entities that were introduced with a buoy are performed with a proform maintained by the non-dominant hand, while lexical elements (with the dominant hand) specify this proform. The structure ends with a blink.Second, the introduction of an entity with a sequential organisation: the succession of a proform and lexical elements that specify the proform. In the same way, the structure ends with a blink.These two structures can be combined.Finally, proforms without a manual referent. These structures contain a proform, but rather than a manual element that specifies it, there is some mouthing or mouth gestures. As with the two previous structures, this one ends with a blink.Using proforms to introduce a new entity is not a discovery. But finding recurrent structures could be a good start with the prospect of formalisation for automatic processing of signed languages.Nevertheless, before any modelling, questions remain unanswered. On the one hand, we started from the premise that buoy structures could be considered as the manifestation of an anaphoric relationship between the proform and the specifier, but this can still be discussed. On the other hand, sequential structures offer less doubt about an analysis in favor of an anaphoric relationship. For proforms with a non-manual and simultaneous specifier, this also raises questions.If we end this article with more questions than precise answers, we think that finding several structures implementing various articulators in a systematic way could eventually allow automatic identification and/or generation of these structures. On the other hand, if identification of a referent could be done more easily, it could help the automatic processing of the reference, if we admit that there is reference (especially for the case of tags).Obviously, this is the first step of an observation that will have to be refined and validated on a larger corpus. But if a generalisation of these structures is possible, then we will be able to define patterns usable in SL processing.; Cet article relate une étude qui porte sur l'analyse linguistique de certains proformes à partir de l’annotation d’un corpus de Langue des Signes Française (LSF) nommé MOCAP1. Nous nous sommes plus particulièrement intéressés aux proformes qui servent à introduire un référent du discours, et nous avons analysé trois structures mettant en œuvre respectivement le maintien d’une configuration avec la main dominée, la séquentialité et l’utilisation de spécifieurs non manuels tels que la labialisation. Nous avons tenté de décrire les formes récurrentes observées pour ces trois structures avec le formalisme AZEE qui prend en compte les particularités des langues signées. Ce même formalisme a également pour vocation d’être utilisé pour le traitement automatique des langues des signes.

http://www.projet-rhapsodie.fr/

In this paper, we show that building a treebank can be used as a way to establish a language. Annotated corpus can be used as tools when arguing that some linguistic data belongs to a separate language (rather than a dialect or variety of another established language). We provide here a case study on a treebank of Naija, a Post-creole spoken in Nigeria which presents us with significant differences from treebanks of English in terms of existing constructions and frequency of several syntactic units.

International audience; This paper discusses how to analyze syntactically irregular expressions in a syntactic treebank. We distinguish such Multi-Word Expressions (MWEs) from comparable non-compositional expressions, i.e. idioms. A solution is proposed in the framework of Universal Dependencies (UD). We further discuss the case of functional MWEs, which are particularly problematic in UD.

International audience; The aim of this paper is to study some characteristics of dependency flux, that is the set of dependencies linking a word on the left with a word on the right in a given position. Based on an exploration of the whole set of UD tree-banks (12M word corpus), we show that what we have called the flux weight, which measures center embeddings, is less than 3 in 99.62 % of the inter-word positions and is bounded by 6, which could be due to short-term memory limitations.

We present three dependency annotation schemes applied to the same corpus of spoken French: Rhapsodie, Orféo and UD (Universal Dependencies). The first two are distributed, the latter is work in progress. We emphasize the annotation of “macrosyntactic” phenomena, i.e.relations in the utterance that are not part of the sub-categorization. We contrast the three schemes and propose a fourth scheme that subsumes the three others., Nous présentons trois schémas d’annotation appliqués à un même corpus de français oral : Rhapsodie, Orféo et UD (Universal Dependencies). Les deux premiers sont diffusés et le troisième est en cours. Nous mettons ici l’accent s ur la macrosyntaxe, c’est-à-dire sur les relations au sein d’un énoncé qui ne relèvent pas de la rection. Nous contrastons les trois schémas et proposons un quatrième schéma qui subsume les trois autres.

International audience

International audience

Our purpose is to allow non-specialists like language learners to access corpora, making syntactic queries easy to express. We propose a methodology including a syntactic parser and using common similarity measures to compare sequences of (automatically produced) morphosyntactic tags., Notre objectif est de permettre aux apprenants de langue d’accéder aux données présentes dans un corpus en facilitant la formulation de requêtes portant sur des critères syntaxiques. Nous proposons une méthodologie utilisant des mesures de similarité classiques pour comparer des séquences d’étiquettes obtenues par des programmes d’annotation.

International audience; We show how to turn a large-scale syntactic dictionary into a dependency-based unification grammar where each piece of lexical information calls a separate rule, yielding a super granular grammar. Subcategorization, raising and control verbs, auxiliaries and copula, passivization, and tough-movement are discussed. We focus on the semantics-syntax interface and offer a new perspective on syntactic structure.

International audience; Using queries to explore corpora is today routine practice not only among researchers in various fields with an empirical approach to discourse, but also among non-specialists who use search engines or con-cordancers for language learning purposes. While keyword-based queries are quite common , non-specialists are less likely to explore syntactic constructions. Syntax-based queries usually require the use of regular expressions with grammatical words combined with mor-phosyntactic tags, meaning that users need to master both the query language of the tool and the tagset of the annotated corpus. However , non-specialists such as language learners may prefer to focus on the output rather than spend time and efforts mastering a query language. To address this shortcoming, we propose a methodology including a syntactic parser and using common similarity measures to compare sequences of automatically produced morphosyntactic tags.

International audience; This paper investigates the POS classification of French words that are called, a.o., « interjections », « sentences-words » (Tesnière), « discursive markers » (in pragmatics) or « inserts » (Biber). Among this kind of words, lexemes such as oui 'yes' and voilà 'there is…' belong to the class that comprehends finite verbs as archetypal members. This class can be labelled predicative. Other lexemes, mainly those generally labeled as interjections or inserts belong to a class named locutive. From a macrosyntactic perspective (Blanche-Benveniste et al.), members of this class are the main units in the structure of the utterance and they convey an intrinsic illocutionary point (Austin). The syntactic properties of locutive words are investigated in a perspective inherited from the transfer mechanism introduced by Tesnière. It appears that interrogative and exclamative pronouns and conjunctions are actually able to transfer units into locutive.; Cet article concerne la classification au sein des parties du discours du français contemporain de mots qui trouvent des appellations diverses selon les approches ; p.ex. « interjections », « mots-phrases » (Tesnière), « marqueurs discursifs » (dans les études sémantico-pragmatiques) ou encore « inserts » (Biber). Les mots phrases tels que oui ou voilà appartiennent à une classe dont les formes verbales indicatives sont les représentants canoniques et que nous nommons prédicatifs, tandis que les autres, principalement les interjections et les inserts, appartiennent à une classe que nous nommons locutifs. Ces derniers forment des noyaux au niveau macrosyntaxique (Blanche-Benveniste et al.) et portent une visée illocutoire intrinsèque (Austin). L'intégration des locutifs au niveau de la syntaxe rectionnelle, dans le système de la translation de Tesnière est étudiée, en montrant notamment que les conjonctions ou pronoms interrogatifs et exclamatifs sont autant de translatifs en locutifs.

International audience

International audience

International audience

International audience

International audience

International audience; This paper proposes a new dependency-based analysis of coordination that generalizes over existing analyses by combining symmetrical and asymmetrical analyses of coordination into a DAG structure. The new joint structure is shown to be theoretically grounded in the notion of connections between words just as the formal definition of other types of dependencies. Beside formalizations of shared dependents (including right-node raising), paradigmatic adverbs, and embedded coordinations , a completely new formalization of non-constituent coordination is proposed.

International audience; Dans cette contribution, nous montrons sur quelles bases biographiques il semblerait que la pensée de Lucien Tesnière se soit développée. Considéré comme l'un des pionniers de la syntaxe dépendancielle, Tesnière invente la syntaxe structurale dans les années trente. Ses écrits de l'époque sont ceux d'un savant ayant l'impression d'avoir eu une véritable « révélation ». Nous montrons ici qu'il faut relativiser cette perception : le grand mérite de Tesnière est d'avoir réussi à agréger des influences diverses dont il n'avait alors pas nécessairement conscience, pour leur donner corps en une théorie d'une remarquable cohérence. Notre étude s'appuie sur les écrits publiés de Tesnière, ainsi que sur l'exploration du fonds Tesnière (BNF NAF 28026), déposé à la Bibliothèque nationale de France par Marie-Hélène Tesnière 1 , qui contient 69 boîtes et fichiers riches en brouillons et correspondances inédits 2. Nous allons voir comment ces influences se sont manifestées au cours de sa vie. Après une brève introduction (1), nous nous focaliserons successivement sur trois aspects de sa carrière : sa vocation en tant qu'enseignant du français aux étrangers (2), sa position de rupture par rapport à ses maîtres (3) et, en bout de course et comme résultante, la fameuse « révélation » qui a fait naître la syntaxe structurale (4).

We show that every set of productions of every natural language can be described as a three-dimensional space: the syntagmatic axis is the axis of combinations of signs within the text, the paradigmatic axis is the axis of commutations in each position in the text, and the semiotic axis is the axis internal to the signs, that links the signified to the signifier, the meaning to the text. Every formal modeling of a natural language has to cover this 3dimensional space. Moreover, in order to capture the regularities, a model seeks after covering the largest possible parts of this space with each rule, which is difficult due to the mutual dependency of the three dimensions. In order to better understand the issues, we will compare two strategies illustrated by a well-known model, LTAG, and another similar one, MTUG, which is inspired by the former but favors a cutting along the syntagmatic and semiotic axes., Nous montrons que tout ensemble de productions d'une langue peut être appréhendé comme un espace à trois dimensions : l'axe syntagmatique est celui des combinaisons de signes au sein du texte, l'axe paradigmatique est celui des commutations possibles en chaque point du texte et l'axe sémiotique est celui interne aux signes, qui lie le signifié au signifiant, le sens au texte. Toute modélisation formelle d'une langue se doit d'effectuer un recouvrement de cet espace tridimensionnel. De plus, pour capter les régularités de la langue, un modèle cherchera à couvrir les portions les plus importantes possibles de cet espace avec chaque règle, ce qui est complexe en raison de l'interdépendance des trois dimensions. Pour mieux comprendre les enjeux, nous comparerons deux stratégies illustrées par un modèle bien connu, LTAG, et un autre similaire, GUST, qui s'en inspire tout en privilégiant un découpage selon les axes syntagmatique et sémiotique.

International audience; This paper scrutinizes various dependency based representations of the syntax of function words, such as prepositions. The focus is on the underlying formal object used to encode the linguistic analyses and its relation to the corresponding linguistic theory. The poly-graph structure is introduced: it consists of a generalization of the concept of graph that allows edges to be vertices of other edges. Such a structure is used to encode dependency-based analyses that are founded on two kinds of morphosyn-tactic criteria: presence constraints and distributional constraints.

In this paper we exhibit macrosyntactic and prosodic fea- tures that are characteristic not only of general discourse genres, but also of more specific types. In our quantitative study, we compare two methods for supervised classification: decision trees and SVM (support vector ma- chines). Furthermore, we address the issue of graphical representations of features through a PCA (principal component analysis). Four main con- clusions can be drawn: (i) in terms of performance , the support vector machines are better suited to our data than decision trees, (ii) certain situa- tional variables are better discriminated than others due to the granularity chosen for analysis, (iii) the principal component analysis is a good way for guessing the complementarity vs. redundancy of the features. All these points provide a valuable feedback on the role played by the intonosyntactic interface in the identification of discourse genres in spoken French., Dans cet article, nous présentons un résumé de certains aspects méthodolo- giques et statistique de notre récent travail sur un corpus de français parlé. Pourquoi compter? Nous montrons comment une étude basée sur des don- nées linguistiques de l'oral, peut avoir une visée technique, où l'objectif se- rait par exemple d'obtenir des erreurs de classification faibles, ou bien avoir une visée plutôt interprétative, où l'objectif serait d'étudier les relations entre niveaux linguistique afin d'en tirer des conclusions. Comment comp- ter? En fonction de cette distinction, on montre que certaines méthodes sta- tistiques sont parfois plus ou moins adéquates. Par exemple, les méthodes de classification automatique de machine learning s'avèrent souvent très efficaces, mais peuvent se révéler beaucoup moins satisfaisante lorsqu'il s'agit de comprendre et d'expliquer l'influence d'une variable linguistique sur une autre. De la même manière, nous montrons qu'une analyse statis- tique peut être utile pour évaluer et même réduire la tâche d'annotation, en permettant par exemple l'identification des variables redondantes qui pour- raient alors être évitées. La discussion est basée sur de nombreux exemples tirés de notre récent travail sur le corpus Rhapsodie, qui est un corpus de français ordinaire, annoté en syntaxe, discours et en prosodie.

International audience; The main objective of the Rhapsodie project (ANR Rhapsodie 07 Corp-030-01) was to define rich, explicit, and reproducible schemes for the annotation of prosody and syntax in different genres (± spontaneous, ± planned, face-to-face interviews vs. broadcast, etc.), in order to study the prosody/syntax/discourse interface in spoken French, and their roles in the segmentation of speech into discourse units (Lacheret, Kahane, & Pietrandrea forthcoming). We here describe the deliverable, a syntactic and prosodic treebank of spoken French, composed of 57 short samples of spoken French (5 minutes long on average, amounting to 3 hours of speech and 33000 words), orthographically and phonetically transcribed. The transcriptions and the annotations are all aligned on the speech signal: phonemes, syllables, words, speakers, overlaps. This resource is freely available at www.projet-rhapsodie.fr. The sound samples (wav/mp3), the acoustic analysis (original F0 curve manually corrected and automatic stylized F0, pitch format), the orthographic transcriptions (txt), the microsyntactic annotations (tabular format), the macrosyntactic annotations (txt, tabular format), the prosodic annotations (xml, textgrid, tabular format), and the metadata (xml and html) can be freely downloaded under the terms of the Creative Commons licence Attribution - Noncommercial - Share Alike 3.0 France. The metadata are encoded in the IMDI-CMFI format and can be parsed on line.

International audience; The aim of this paper is to describe an automated process to segment spoken French transcribed data into macrosyntactic units. While sentences are delimited by punctuation marks for written data, there is no obvious hint nor limit to major units for speech. As a reference, we used the manual annotation of macrosyntactic units based on illocutionary as well as syntactic criteria and developed for the Rhapsodie corpus, a 33.000 words prosodic and syntactic treebank. Our segmenters were built using machine learning methods as supervised classifiers~: segmentation is about identifying the boundaries of units, which amounts to classifying each interword space. We trained six different models on Rhapsodie using different sets of features, including prosodic and morphosyntactic cues, on the assumption that their combination would be relevant for the task. Both types of cues could be resulting either from manual annotation/correction or from fully automated processes, which comparison might help determine the cost of manual effort, especially for the 3M words of spoken French of the Orfeo project those experiments are contributing to.

International audience; The aim of this paper is to describe an automated process to segment spoken French transcribed data into macrosyntactic units. While sentences are delimited by punctuation marks for written data, there is no obvious hint nor limit to major units for speech. As a reference, we used the manual annotation of macrosyntactic units based on illocutionary as well as syntactic criteria and developed for the Rhapsodie corpus, a 33.000 words prosodic and syntactic treebank. Our segmenters were built using machine learning methods as supervised classifiers : segmentation is about identifying the boundaries of units, which amounts to classifying each interword space. We trained six different models on Rhapsodie using different sets of features, including prosodic and morphosyntactic cues, on the assumption that their combination would be relevant for the task. Both types of cues could be resulting either from manual annotation/correction or from fully automated processes, which comparison might help determine the cost of manual effort, especially for the 3M words of spoken French of the Orfeo project those experiments are contributing to.

International audience; This article presents the methods, results, and precision of the syntactic annotation process of the Rhapsodie Treebank of spoken French. The Rhapsodie Treebank is an 33,000 word corpus annotated for prosody and syntax, licensed in its entirety under Creative Commons. The syntactic annotation contains two levels: a macro-syntactic level, containing a segmentation into illocutionary units (including discourse markers, parentheses ...) and a micro-syntactic level including dependency relations and various paradigmatic structures, called pile constructions, the latter being particularly frequent and diverse in spoken language. The micro-syntactic annotation process, presented in this paper, includes a semi-automatic preparation of the transcription, the application of a syntactic dependency parser, transcoding of the parsing results to the Rhapsodie annotation scheme, manual correction by multiple annotators followed by a validation process, and finally the application of coherence rules that check common errors. The good inter-annotator agreement scores are presented and analyzed in greater detail. The article also includes the list of functions used in the dependency annotation and for the distinction of various pile constructions and presents the ideas underlying these choices.