The Winograd Schema Challenge (WSC) consists of a set of anaphora resolution problems resolvable only by reasoning about world knowledge. This article describes the update of the existing French data set and the creation of three subsets allowing for a more robust, fine-grained evaluation protocol of WSC in French (FWSC) : an associative subset (items easily resolvable with lexical co-occurrence), a switchable subset (items where the inversion of two keywords reverses the answer) and a negatable subset (items where applying negation on its verb reverses the answer). Experiences on these data sets with CamemBERT reach SOTA performances. Our evaluation protocol showed in addition that the higher performance could be explained by the existence of associative items in FWSC. Besides, increasing the size of training corpus improves the model’s performance on switchable items while the impact of larger training corpus remains small on negatable items.

Nous présentons un résumé en français et un résumé en anglais de l’article Is Attention Explanation ? An Introduction to the Debate (Bibal et al., 2022), publié dans les actes de la conférence 60th Annual Meeting of the Association for Computational Linguistics (ACL 2022).

Mastering a foreign language like English can bring better opportunities. In this context, although multiword expressions (MWE) are associated with proficiency, they are usually neglected in the works of automatic scoring language learners. Therefore, we study MWE-based features (i.e., occurrence and concreteness) in this work, aiming at assessing their relevance for automated essay scoring. To achieve this goal, we also compare MWE features with other classic features, such as length-based, graded resource, orthographic neighbors, part-of-speech, morphology, dependency relations, verb tense, language development, and coherence. Although the results indicate that classic features are more significant than MWE for automatic scoring, we observed encouraging results when looking at the MWE concreteness through the levels.

The performance of deep learning models in NLP and other fields of machine learning has led to a rise in their popularity, and so the need for explanations of these models becomes paramount. Attention has been seen as a solution to increase performance, while providing some explanations. However, a debate has started to cast doubt on the explanatory power of attention in neural networks. Although the debate has created a vast literature thanks to contributions from various areas, the lack of communication is becoming more and more tangible. In this paper, we provide a clear overview of the insights on the debate by critically confronting works from these different areas. This holistic vision can be of great interest for future works in all the communities concerned by this debate. We sum up the main challenges spotted in these areas, and we conclude by discussing the most promising future avenues on attention as an explanation.

In this paper, we present the FABRA: readability toolkit based on the aggregation of a large number of readability predictor variables. The toolkit is implemented as a service-oriented architecture, which obviates the need for installation, and simplifies its integration into other projects. We also perform a set of experiments to show which features are most predictive on two different corpora, and how the use of aggregators improves performance over standard feature-based readability prediction. Our experiments show that, for the explored corpora, the most important predictors for native texts are measures of lexical diversity, dependency counts and text coherence, while the most important predictors for foreign texts are syntactic variables illustrating language development, as well as features linked to lexical sophistication. FABRA: have the potential to support new research on readability assessment for French.

Cet article décrit la participation de l’équipe Nantalco à la tâche 2 du Défi Fouille de Textes 2021 (DEFT) : évaluation automatique de copies d’après une référence existante. Nous avons utilisé principalement des traits basés sur la similarité cosinus des deux vecteurs représentant la similarité textuelle entre des réponses d’étudiant et la référence. Plusieurs types de vecteurs ont été utilisés (vecteur d’occurrences de mots, vecteur tf-idf, embeddings non contextualisés de fastText, embeddings contextualisés de CamemBERT et enfin Sentence Embeddings Multilingues ajustés sur des corpus multilingues). La meilleure performance du concours sur cette tâche a été de 0.682 (précision) et celle de notre équipe 0.639. Cette performance a été obtenue avec les Sentence Embeddings Multilingues alors que celle des embeddings non ajustés ne s’est élevée qu’à 0.55, suggérant que de récents modèles de langues pré-entraînés doivent être fine-tunés afin d’avoir des embeddings adéquats au niveau phrastique.