BERT-like language models have been demonstrated to capture the idiomatic meaning of multiword expressions. Linguists have also shown that idioms have varying degrees of idiomaticity. In this paper, we assess CamemBERT’s sensitivity to the degree of idiomaticity within idioms, as well as the dependency of this sensitivity on part of speech and idiom length. We used a demasking task on tokens from 3127 idioms and 22551 tokens corresponding to simple lexemes taken from the French Lexical Network (LN-fr), and observed that CamemBERT performs distinctly on tokens embedded within idioms compared to simple ones. When demasking tokens within idioms, the model is not proficient in discerning their level of idiomaticity. Moreover, regardless of idiomaticity, CamemBERT excels at handling function words. The length of idioms also impacts CamemBERT’s performance to a certain extent. The last two observations partly explain the difference between the model’s performance on idioms versus simple lexemes. We conclude that the model treats idioms differently from simple lexemes, but that it does not capture the difference in compositionality between subclasses of idioms.

GenDR is a text realizer that takes as input a graph-based semantic representation and outputs the corresponding syntactic dependency trees. One of the tasks in this transduction is lexicalization, i.e., choosing the right lexical units to express a given semanteme. To do so, GenDR uses a semantic dictionary that maps semantemes to corresponding lexical units in a given language. This study aims to develop a flexible lexicalization module to automatically build a rich semantic dictionary for French. To achieve this, we tried two methods. The first one consisted in extracting information from the French Lexical Network, a large-scale French lexical resource, and adapting it to GenDR. The second one was to test a contextual neural language model’s ability to generate potential additional lexicalizations. The first method significantly broadened the coverage of GenDR, while the additional lexicalizations produced by the language model turned out to be of limited use, which brings us to the conclusion that it is not suited to perform the task we’ve asked from it.

Rule-based text generators lack the coverage and fluency of their neural counterparts, but have two big advantages over them: (i) they are entirely controllable and do not hallucinate; and (ii) they can fully explain how an output was generated from an input. In this paper we leverage these two advantages to create large and reliable synthetic datasets with multiple human-intelligible intermediate representations. We present the Modular Data-to-Text (Mod-D2T) Dataset which incorporates ten intermediate-level representations between input triple sets and output text; the mappings from one level to the next can broadly be interpreted as the traditional modular tasks of an NLG pipeline. We describe the Mod-D2T dataset, evaluate its quality via manual validation and discuss its applications and limitations. Data, code and documentation are available at https://github.com/mille-s/Mod-D2T.

Nodes in Abstract Meaning Representation (AMR) are generally thought of as neo-Davidsonian entities. We review existing translation into neo-Davidsonian representations and show that these translations inconsistently handle copula sentences. We link the problem to an asymmetry arising from a problematic handling of words with no associated PropBank frames for the underlying predicate. We introduce a method to automatically and uniformly decompose AMR nodes into an entity-part and a predicative part, which offers a consistent treatment of copula sentences and quasi- predicates such as brother or client.

Different algorithms have been proposed to detect semantic shifts (changes in a word meaning over time) in a diachronic corpus. Yet, and somehow surprisingly, no reference corpus has been designed so far to evaluate them, leaving researchers to fallback to troublesome evaluation strategies. In this work, we introduce a methodology for the construction of a reference dataset for the evaluation of semantic shift detection, that is, a list of words where we know for sure whether they present a word meaning change over a period of interest. We leverage a state-of-the-art word-sense disambiguation model to associate a date of first appearance to all the senses of a word. Significant changes in sense distributions as well as clear stability are detected and the resulting words are inspected by experts using a dedicated interface before populating a reference dataset. As a proof of concept, we apply this methodology to a corpus of newspapers from Quebec covering the whole 20th century. We manually verified a subset of candidates, leading to QC-FR-Diac-V1.0, a corpus of 151 words allowing one to evaluate the identification of semantic shifts in French between 1910 and 1990.

While idioms are usually very rigid in their expression, they sometimes allow a certain level of freedom in their usage, with modifiers or complements splitting them or being syntactically attached to internal nodes rather than to the root (e.g., “take something with a big grain of salt”). This means that they cannot always be handled as ready-made strings in rule-based natural language generation systems. Having access to the internal syntactic structure of an idiom allows for more subtle processing. We propose a way to enumerate all possible language-independent n-node trees and to map particular idioms of a language onto these generic syntactic patterns. Using this method, we integrate the idioms from the LN-fr into GenDR, a multilingual realizer. Our implementation covers nearly 98% of LN-fr’s idioms with high precision, and can easily be extended or ported to other languages.

We propose a multilingual method for the extraction of biased sentences from Wikipedia, and use it to create corpora in Bulgarian, French and English. Sifting through the revision history of the articles that at some point had been considered biased and later corrected, we retrieve the last tagged and the first untagged revisions as the before/after snapshots of what was deemed a violation of Wikipedia’s neutral point of view policy. We extract the sentences that were removed or rewritten in that edit. The approach yields sufficient data even in the case of relatively small Wikipedias, such as the Bulgarian one, where 62k articles produced 5k biased sentences. We evaluate our method by manually annotating 520 sentences for Bulgarian and French, and 744 for English. We assess the level of noise and analyze its sources. Finally, we exploit the data with well-known classification methods to detect biased sentences. Code and datasets are hosted at https://github.com/crim-ca/wiki-bias.

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.

A lexical function represents a type of relation that exists between lexical units (words or expressions) in any language. For example, the antonymy is a type of relation that is represented by the lexical function Anti: Anti(big) = small. Those relations include both paradigmatic relations, i.e. vertical relations, such as synonymy, antonymy and meronymy and syntagmatic relations, i.e. horizontal relations, such as objective qualification (legitimate demand), subjective qualification (fruitful analysis), positive evaluation (good review) and support verbs (pay a visit, subject to an interrogation). In this paper, we present the Lexical Functions Ontology Model (lexfom) to represent lexical functions and the relation among lexical units. Lexfom is divided in four modules: lexical function representation (lfrep), lexical function family (lffam), lexical function semantic perspective (lfsem) and lexical function relations (lfrel). Moreover, we show how it combines to Lexical Model for Ontologies (lemon), for the transformation of lexical networks into the semantic web formats. So far, we have implemented 100 simple and 500 complex lexical functions, and encoded about 8,000 syntagmatic and 46,000 paradigmatic relations, for the French language.

En sémantique distributionnelle, le sens des mots est modélisé par des vecteurs qui représentent leur distribution en corpus. Les modèles étant souvent calculés sur des corpus sans pré-traitement linguistique poussé, ils ne permettent pas de rendre bien compte de la compositionnalité morphologique des mots-formes. Nous proposons une méthode pour décomposer les vecteurs de mots en vecteurs lexicaux et flexionnels.

Pour concevoir des générateurs automatiques de texte génériques qui soient facilement réutilisables d’une langue et d’une application à l’autre, il faut modéliser les principaux phénomènes linguistiques qu’on retrouve dans les langues en général. Un des phénomènes fondamentaux qui demeurent problématiques pour le TAL est celui des collocations, comme grippe carabinée, peur bleue ou désir ardent, où un sens (ici, l’intensité) ne s’exprime pas de la même façon selon l’unité lexicale qu’il modifie. Dans la lexicographie explicative et combinatoire, on modélise les collocations au moyen de fonctions lexicales qui correspondent à des patrons récurrents de collocations. Par exemple, les expressions mentionnées ici se décrivent au moyen de la fonction Magn : Magn(PEUR) = BLEUE, Magn(GRIPPE) = CARABINÉE, etc. Il existe des centaines de fonctions lexicales. Dans cet article, nous nous intéressons à l’implémentation d’un sous-ensemble de fonctions qui décrivent les verbes supports et certains types de modificateurs.

Nous proposons une formalisation de la décomposition du sens dans le cadre de la Grammaire d’Unification Sens-Texte. Cette formalisation vise une meilleure intégration des décompositions sémantiques dans un modèle global de la langue. Elle repose sur un jeu de saturation de polarités qui permet de contrôler la construction des représentations décomposées ainsi que leur mise en correspondance avec des arbres syntaxiques qui les expriment. Le formalisme proposé est illustré ici dans une perspective de synthèse, mais il s’applique également en analyse.

L’objectif de cet article est de présenter l’état actuel du modèle de la Grammaire d’Unification Sens-Texte, notamment depuis que les bases formelles du modèle ont été éclaircies grâce au développement des Grammaires d’Unification Polarisées. L’accent est mis sur l’architecture du modèle et le rôle de la polarisation dans l’articulation des différents modules — l’interface sémantique-syntaxe, l’interface syntaxe-morphotopologie et les grammaires décrivant les différents niveaux de représentation. Nous étudions comment les procédures d’analyse et de génération sont contrôlables par différentes stratégies de neutralisation des différentes polarités.