In this paper, we present a study of transformer-based Named Entity Recognition (NER) as applied to Ancient Greek texts, with an emphasis on retrieving personal names. Recent research shows that, while the task remains difficult, the use of transformer models results in significant improvements. We, therefore, compare the performance of four transformer models on the task of NER for the categories of people, locations and groups, and add an out-of-domain test set to the existing datasets. Results on this set highlight the shortcomings of the models when confronted with a random sample of sentences. To be able to more straightforwardly integrate domain and linguistic knowledge to improve performance, we narrow down our approach to the category of people. The task is simplified to a binary PERS/MISC classification on the token level, starting from capitalised words. Next, we test the use of domain and linguistic knowledge to improve the results. We find that including simple gazetteer information as a binary mask has a marginally positive effect on newly annotated data and that treebanks can be used to help identify multi-word individuals if they are scarcely or inconsistently annotated in the available training data. The qualitative error analysis identifies the potential for improvement in both manual annotation and the inclusion of domain and linguistic knowledge in the transformer models.

This paper investigates the efficacy of multilingual models for the task of text-to-AMR parsing, focusing on English, Spanish, and Dutch. We train and evaluate models under various configurations, including monolingual and multilingual settings, both in full and reduced data scenarios. Our empirical results reveal that while monolingual models exhibit superior performance, multilingual models are competitive across all languages, offering a more resource-efficient alternative for training and deployment. Crucially, our findings demonstrate that AMR parsing benefits from transfer learning across languages even when having access to significantly smaller datasets. As a tangible contribution, we provide text-to-AMR parsing models for the aforementioned languages as well as multilingual variants, and make available the large corpora of translated data for Dutch, Spanish (and Irish) that we used for training them in order to foster AMR research in non-English languages. Additionally, we open-source the training code and offer an interactive interface for parsing AMR graphs from text.

In French, the placement of the adjective within a noun phrase is subject to variation: it can appear either before or after the noun. We conduct experiments to assess whether transformer-based language models are able to learn the adjective position in noun phrases in French –a position which depends on several linguistic factors. Prior findings have shown that transformer models are insensitive to permutated word order, but in this work, we show that finetuned models are successful at learning and selecting the correct position of the adjective. However, this success can be attributed to the process of finetuning rather than the linguistic knowledge acquired during pretraining, as evidenced by the low accuracy of experiments of classification that make use of pretrained embeddings. Comparing the finetuned models to the choices of native speakers (with a questionnaire), we notice that the models favor context and global syntactic roles, and are weaker with complex structures and fixed expressions.

We evaluate the performance of various models on the task of named entity recognition (NER) for classical Latin. Using an existing dataset, we train two transformer-based LatinBERT models and one shallow conditional random field (CRF) model. The performance is assessed using both standard metrics and a detailed manual error analysis, and compared to the results obtained by different already released Latin NER tools. Both analyses demonstrate that the BERT models achieve a better f1-score than the other models. Furthermore, we annotate new, unseen data for further evaluation of the models, and we discuss the impact of annotation choices on the results.

Aspect is a linguistic concept that describes how an action, event, or state of a verb phrase is situated in time. In this paper, we explore whether different transformer models are capable of identifying aspectual features. We focus on two specific aspectual features: telicity and duration. Telicity marks whether the verb’s action or state has an endpoint or not (telic/atelic), and duration denotes whether a verb expresses an action (dynamic) or a state (stative). These features are integral to the interpretation of natural language, but also hard to annotate and identify with NLP methods. We perform experiments in English and French, and our results show that transformer models adequately capture information on telicity and duration in their vectors, even in their non-finetuned forms, but are somewhat biased with regard to verb tense and word order.

CLIP, a vision-language network trained with a multimodal contrastive learning objective on a large dataset of images and captions, has demonstrated impressive zero-shot ability in various tasks. However, recent work showed that in comparison to unimodal (visual) networks, CLIP’s multimodal training does not benefit generalization (e.g. few-shot or transfer learning) for standard visual classification tasks such as object, street numbers or animal recognition. Here, we hypothesize that CLIP’s improved unimodal generalization abilities may be most prominent in domains that involve human-centric concepts (cultural, social, aesthetic, affective...); this is because CLIP’s training dataset is mainly composed of image annotations made by humans for other humans. To evaluate this, we use 3 tasks that require judging human-centric concepts”:” sentiment analysis on tweets, genre classification on books or movies. We introduce and publicly release a new multimodal dataset for movie genre classification. We compare CLIP’s visual stream against two visually trained networks and CLIP’s textual stream against two linguistically trained networks, as well as multimodal combinations of these networks. We show that CLIP generally outperforms other networks, whether using one or two modalities. We conclude that CLIP’s multimodal training is beneficial for both unimodal and multimodal tasks that require classification of human-centric concepts.

New models for natural language understanding have recently made an unparalleled amount of progress, which has led some researchers to suggest that the models induce universal text representations. However, current benchmarks are predominantly targeting semantic phenomena; we make the case that pragmatics needs to take center stage in the evaluation of natural language understanding. We introduce PragmEval, a new benchmark for the evaluation of natural language understanding, that unites 11 pragmatics-focused evaluation datasets for English. PragmEval can be used as supplementary training data in a multi-task learning setup, and is publicly available, alongside the code for gathering and preprocessing the datasets. Using our evaluation suite, we show that natural language inference, a widely used pretraining task, does not result in genuinely universal representations, which presents a new challenge for multi-task learning.

De nombreuses tâches sémantiques en TAL font usage de données collectées de manière semiautomatique, ce qui est souvent source d’artefacts indésirables qui peuvent affecter négativement les modèles entraînés sur celles-ci. Avec l’évolution plus récente vers des modèles à usage générique pré-entraînés plus complexes, et moins interprétables, ces biais peuvent conduire à l’intégration de corrélations indésirables dans des applications utilisateurs. Récemment, quelques méthodes ont été proposées pour entraîner des plongements de mots avec une meilleure interprétabilité. Nous proposons une méthode simple qui exploite ces représentations pour détecter de manière préventive des corrélations lexicales faciles à apprendre, dans divers jeux de données. Nous évaluons à cette fin quelques modèles de plongements interprétables populaires pour l’anglais, en utilisant à la fois une évaluation intrinsèque, et un ensemble de tâches sémantiques en aval, et nous utilisons la qualité interprétable des plongements afin de diagnostiquer des biais potentiels dans les jeux de données associés.

L’aspect du verbe décrit la manière dont une action, un événement ou un état exprimé par un verbe est lié au temps ; la télicité est la propriété d’un syntagme verbal qui présente une action ou un événement comme étant mené à son terme ; la durée distingue les verbes qui expriment une action (dynamique) ou un état (statique). Ces caractéristiques essentielles à l’interprétation du langage naturel, sont également difficiles à annoter et à identifier par les méthodes de TAL. Dans ce travail, nous estimons la capacité de différents modèles de type transformers pré-entraînés (BERT, RoBERTa, XLNet, ALBERT) à prédire la télicité et la durée. Nos résultats montrent que BERT est le plus performant sur les deux tâches, tandis que les modèles XLNet et ALBERT sont les plus faibles. Par ailleurs, les performances de la plupart des modèles sont améliorées lorsqu’on leur fournit en plus la position des verbes. Globalement, notre étude établit que les modèles de type transformers captent en grande partie la télicité et la durée.

Selectional preference is defined as the tendency of a predicate to favor particular arguments within a certain linguistic context, and likewise, reject others that result in conflicting or implausible meanings. The stellar success of contextual word embedding models such as BERT in NLP tasks has led many to question whether these models have learned linguistic information, but up till now, most research has focused on syntactic information. We investigate whether Bert contains information on the selectional preferences of words, by examining the probability it assigns to the dependent word given the presence of a head word in a sentence. We are using word pairs of head-dependent words in five different syntactic relations from the SP-10K corpus of selectional preference (Zhang et al., 2019b), in sentences from the ukWaC corpus, and we are calculating the correlation of the plausibility score (from SP-10K) and the model probabilities. Our results show that overall, there is no strong positive or negative correlation in any syntactic relation, but we do find that certain head words have a strong correlation and that masking all words but the head word yields the most positive correlations in most scenarios –which indicates that the semantics of the predicate is indeed an integral and influential factor for the selection of the argument.

Using a model trained to predict discourse markers between sentence pairs, we predict plausible markers between sentence pairs with a known semantic relation (provided by existing classification datasets). These predictions allow us to study the link between discourse markers and the semantic relations annotated in classification datasets. Handcrafted mappings have been proposed between markers and discourse relations on a limited set of markers and a limited set of categories, but there exists hundreds of discourse markers expressing a wide variety of relations, and there is no consensus on the taxonomy of relations between competing discourse theories (which are largely built in a top-down fashion). By using an automatic prediction method over existing semantically annotated datasets, we provide a bottom-up characterization of discourse markers in English. The resulting dataset, named DiscSense, is publicly available.

In the last few years, a number of successful approaches have emerged that are able to adequately model various aspects of natural language. In particular, language models based on neural networks have improved the state of the art with regard to predictive language modeling, while topic models are successful at capturing clear-cut, semantic dimensions. In this paper, we will explore how these approaches can be adapted and combined to model the linguistic and literary aspects needed for poetry generation. The system is exclusively trained on standard, non-poetic text, and its output is constrained in order to confer a poetic character to the generated verse. The framework is applied to the generation of poems in both English and French, and is equally evaluated for both languages. Even though it only uses standard, non-poetic text as input, the system yields state of the art results for poetry generation.

Various NLP problems – such as the prediction of sentence similarity, entailment, and discourse relations – are all instances of the same general task: the modeling of semantic relations between a pair of textual elements. A popular model for such problems is to embed sentences into fixed size vectors, and use composition functions (e.g. concatenation or sum) of those vectors as features for the prediction. At the same time, composition of embeddings has been a main focus within the field of Statistical Relational Learning (SRL) whose goal is to predict relations between entities (typically from knowledge base triples). In this article, we show that previous work on relation prediction between texts implicitly uses compositions from baseline SRL models. We show that such compositions are not expressive enough for several tasks (e.g. natural language inference). We build on recent SRL models to address textual relational problems, showing that they are more expressive, and can alleviate issues from simpler compositions. The resulting models significantly improve the state of the art in both transferable sentence representation learning and relation prediction.

La génération automatique de poésie est une tâche ardue pour un système informatique. Pour qu’un poème ait du sens, il est important de prendre en compte à la fois des aspects linguistiques et littéraires. Ces dernières années, un certain nombre d’approches fructueuses sont apparues, capables de modéliser de manière adéquate divers aspects du langage naturel. En particulier, les modèles de langue basés sur les réseaux de neurones ont amélioré l’état de l’art par rapport à la modélisation prédictive de langage, tandis que les topic models sont capables de capturer une certaine cohérence thématique. Dans cet article, on explorera comment ces approches peuvent être adaptées et combinées afin de modéliser les aspects linguistiques et littéraires nécessaires pour la génération de poésie. Le système est exclusivement entraîné sur des textes génériques, et sa sortie est contrainte afin de conférer un caractère poétique au vers généré. Le cadre présenté est appliqué à la génération de poèmes en français, et évalué à l’aide d’une évaluation humaine.

Nous présentons le système utilisé par l’équipe Synapse/IRIT dans la compétition DEFT2019 portant sur deux tâches liées à des cas cliniques rédigés en français : l’une d’appariement entre des cas cliniques et des discussions, l’autre d’extraction de mots-clefs. Une des particularité est l’emploi d’apprentissage non-supervisé sur les deux tâches, sur un corpus construit spécifiquement pour le domaine médical en français

Current state of the art systems in NLP heavily rely on manually annotated datasets, which are expensive to construct. Very little work adequately exploits unannotated data – such as discourse markers between sentences – mainly because of data sparseness and ineffective extraction methods. In the present work, we propose a method to automatically discover sentence pairs with relevant discourse markers, and apply it to massive amounts of data. Our resulting dataset contains 174 discourse markers with at least 10k examples each, even for rare markers such as “coincidentally” or “amazingly”. We use the resulting data as supervision for learning transferable sentence embeddings. In addition, we show that even though sentence representation learning through prediction of discourse marker yields state of the art results across different transfer tasks, it’s not clear that our models made use of the semantic relation between sentences, thus leaving room for further improvements.

Nous présentons le système utilisé par l’équipe Melodi/Synapse Développement dans la compétition DEFT2018 portant sur la classification de thématique ou de sentiments de tweets en français. On propose un système unique pour les deux approches qui combine concaténativement deux méthodes d’embedding et trois modèles de représentation séquence. Le système se classe 1/13 en analyse de sentiments et 4/13 en classification thématique.

Plusieurs tâches en traitement du langage naturel impliquent de modifier des phrases en conservant au mieux leur sens, comme la reformulation, la compression, la simplification, chacune avec leurs propres données et modèles. Nous introduisons ici une méthode générale s’adressant à tous ces problèmes, utilisant des données plus simples à obtenir : un ensemble de phrases munies d’indicateurs sur leur style, comme des phrases et le type de sentiment qu’elles expriment. Cette méthode repose sur un modèle d’apprentissage de représentations non supervisé (un auto-encodeur variationnel), puis sur le changement des représentations apprises pour correspondre à un style donné. Le résultat est évalué qualitativement, puis quantitativement sur le jeu de données de compression de phrases Microsoft, avec des résultats encourageants.

Automatic acquisition of semantics from text has received quite some attention in natural language processing. A lot of research has been done by looking at syntactically similar contexts. For example, semantically related nouns can be clustered by looking at the collocating adjectives. There are, however, two major problems with this approach : computational complexity and data sparseness. This paper describes the application of a mathematical technique called singular value decomposition, which has been succesfully applied in Information Retrieval to counter these problems. It is investigated whether this technique is also able to cluster nouns according to latent semantic dimensions in a reduced adjective space.