Logical reasoning remains a challenge for natural language processing, but it can be improved by training language models to mimic theorem provers on procedurally generated problems. Previous work used domain-specific proof generation algorithms, which biases reasoning toward specific proof traces and limits auditability and extensibility. We present a simpler and more general declarative framework with flexible context-sensitive rules binding multiple languages (specifically, simplified English and the TPTP theorem-proving language). We construct first-order logic problems by selecting up to 32 premises and one hypothesis. We demonstrate that using semantic constraints during generation and careful English verbalization of predicates enhances logical reasoning without hurting natural English tasks. Using relatively small DeBERTa-v3 models, we achieve state-of-the-art accuracy on the FOLIO human-authored logic dataset, surpassing GPT-4 in accuracy with or without an external solver by 12%.

This paper presents DISRPT, a multilingual, multi-domain, and cross-framework benchmark dataset for discourse processing, covering the tasks of discourse unit segmentation, connective identification, and relation classification. DISRPT includes 13 languages, with data from 24 corpora covering about 4 millions tokens and around 250,000 discourse relation instances from 4 discourse frameworks: RST, SDRT, PDTB, and Discourse Dependencies. We present an overview of the data, its development across three NLP shared tasks on discourse processing carried out in the past five years, and the latest modifications and added extensions. We also carry out an evaluation of state-of-the-art multilingual systems trained on the data for each task, showing plateau performance on segmentation, but important room for improvement for connective identification and relation classification. The DISRPT benchmark employs a unified format that we make available on GitHub and HuggingFace in order to encourage future work on discourse processing across languages, domains, and frameworks.

Medical multiple-choice question answering (MCQA) is a challenging evaluation for medical natural language processing and a helpful task in itself. Medical questions may describe patient symptoms and ask for the correct diagnosis, which requires domain knowledge and complex reasoning. Standard language modeling pretraining alone is not sufficient to achieve the best results with BERT-base size (Devlin et al., 2019) encoders. Jin et al. (2020) showed that focusing masked language modeling on disease name prediction when using medical encyclopedic paragraphs as input leads to considerable MCQA accuracy improvement. In this work, we show that (1) fine-tuning on generated MCQA dataset outperforms the masked language modeling based objective and (2) correctly masking the cues to the answers is critical for good performance. We release new pretraining datasets and achieve state-of-the-art results on 4 MCQA datasets, notably +5.7% with base-size model on MedQA-USMLE.

The HuggingFace Datasets Hub hosts thousands of datasets, offering exciting opportunities for language model training and evaluation. However, datasets for a specific task type often have different structures, making harmonization challenging which prevents the interchangeable use of comparable datasets. As a result, multi-task training or evaluation necessitates manual work to fit data into task templates. Several initiatives independently tackle this issue by releasing harmonized datasets or providing harmonization codes to preprocess datasets into a consistent format. We identify patterns in such preprocessings, such as column renaming, or more complex patterns. We then propose an annotation framework that enables concise, readable, and reusable preprocessing annotations. tasksource annotates more than 600 task preprocessings and provides a backend to automate dataset alignment. We fine-tune a multi-task text encoder on all tasksource tasks, outperforming every publicly available text encoder of comparable parameter count according to an external evaluation.

Theory of Mind (ToM) is a critical component of intelligence but its assessment remains the subject of heated debates. Prior research applied human ToM assessments to natural language processing models using either human-created standardized tests or rule-based templates. However, these methods primarily focus on simplistic reasoning and require further validation. Here, we leverage dynamic epistemic logic to isolate a particular component of ToM and to generate controlled problems. We also introduce new verbalization techniques to express these problems in English natural language. Our findings indicate that some language model scaling (from 70M to 6B and 350M to 174B) does not consistently yield results better than random chance. While GPT-4 demonstrates superior epistemic reasoning capabilities, there is still room for improvement. Our code and datasets are publicly available.

Data augmentation is an important method for evaluating the robustness of and enhancing the diversity of training data for natural language processing (NLP) models. In this paper, we present NL-Augmenter, a new participatory Python-based natural language (NL) augmentation framework which supports the creation of transformations (modifications to the data) and filters (data splits according to specific features). We describe the framework and an initial set of 117 transformations and 23 filters for a variety of NL tasks annotated with noisy descriptive tags. The transformations incorporate noise, intentional and accidental human mistakes, socio-linguistic variation, semantically-valid style, syntax changes, as well as artificial constructs that are unambiguous to humans. We demonstrate the efficacy of NL-Augmenter by using its transformations to analyze the robustness of popular language models. We find different models to be differently challenged on different tasks, with quasi-systematic score decreases. The infrastructure, datacards, and robustness evaluation results are publicly available on GitHub for the benefit of researchers working on paraphrase generation, robustness analysis, and low-resource NLP.

Words of Estimative Probability (WEP) are phrases used to express the plausibility of a statement. Examples include terms like \textit{probably, maybe, likely, doubt, unlikely}, and \textit{impossible}. Surveys have shown that human evaluators tend to agree when assigning numerical probability levels to these WEPs. For instance, the term \textit{highly likely} equates to a median probability of $0.90{\pm}0.08$ according to a survey by \citet{fagen-ulmschneider}.In this study, our focus is to gauge the competency of neural language processing models in accurately capturing the consensual probability level associated with each WEP. Our first approach is utilizing the UNLI dataset \cite{chen-etal-2020-uncertain}, which links premises and hypotheses with their perceived joint probability $p$. From this, we craft prompts in the form: "[\textsc{Premise}]. [\textsc{Wep}], [\textsc{Hypothesis}].” This allows us to evaluate whether language models can predict if the consensual probability level of a WEP aligns closely with $p$.In our second approach, we develop a dataset based on WEP-focused probabilistic reasoning to assess if language models can logically process WEP compositions. For example, given the prompt "[\textsc{EventA}] \textit{is likely}. [\textsc{EventB}] \textit{is impossible}.”, a well-functioning language model should not conclude that [\textsc{EventA$\&$B}] is likely. Through our study, we observe that both tasks present challenges to out-of-the-box English language models. However, we also demonstrate that fine-tuning these models can lead to significant and transferable improvements.

Task embeddings are low-dimensional representations that are trained to capture task properties. In this paper, we propose MetaEval, a collection of 101 NLP tasks. We fit a single transformer to all MetaEval tasks jointly while conditioning it on learned embeddings. The resulting task embeddings enable a novel analysis of the space of tasks. We then show that task aspects can be mapped to task embeddings for new tasks without using any annotated examples. Predicted embeddings can modulate the encoder for zero-shot inference and outperform a zero-shot baseline on GLUE tasks. The provided multitask setup can function as a benchmark for future transfer learning research.

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.

We describe our approach for SemEval-2021 task 6 on detection of persuasion techniques in multimodal content (memes). Our system combines pretrained multimodal models (CLIP) and chained classifiers. Also, we propose to enrich the data by a data augmentation technique. Our submission achieves a rank of 8/16 in terms of F1-micro and 9/16 with F1-macro on the test set.

Visual grounding is a promising path toward more robust and accurate Natural Language Processing (NLP) models. Many multimodal extensions of BERT (e.g., VideoBERT, LXMERT, VL-BERT) allow a joint modeling of texts and images that lead to state-of-the-art results on multimodal tasks such as Visual Question Answering. Here, we leverage multimodal modeling for purely textual tasks (language modeling and classification) with the expectation that the multimodal pretraining provides a grounding that can improve text processing accuracy. We propose possible strategies in this respect. A first type of strategy, referred to as transferred grounding consists in applying multimodal models to text-only tasks using a placeholder to replace image input. The second one, which we call associative grounding, harnesses image retrieval to match texts with related images during both pretraining and text-only downstream tasks. We draw further distinctions into both strategies and then compare them according to their impact on language modeling and commonsense-related downstream tasks, showing improvement over text-only baselines.

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.

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.

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.