Itziar Gonzalez-Dios

BigScience Workshop, Teven Le Scao, Angela Fan et al. · allen-ai, berkeley

Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and applications using LLMs, we publicly release our models and code under the Responsible AI License.

Hugo Laurençon, Lucile Saulnier, Thomas Wang et al. · huggingface

As language models grow ever larger, the need for large-scale high-quality text datasets has never been more pressing, especially in multilingual settings. The BigScience workshop, a 1-year international and multidisciplinary initiative, was formed with the goal of researching and training large language models as a values-driven undertaking, putting issues of ethics, harm, and governance in the foreground. This paper documents the data creation and curation efforts undertaken by BigScience to assemble the Responsible Open-science Open-collaboration Text Sources (ROOTS) corpus, a 1.6TB dataset spanning 59 languages that was used to train the 176-billion-parameter BigScience Large Open-science Open-access Multilingual (BLOOM) language model. We further release a large initial subset of the corpus and analyses thereof, and hope to empower large-scale monolingual and multilingual modeling projects with both the data and the processing tools, as well as stimulate research around this large multilingual corpus.

Oscar Sainz, Itziar Gonzalez-Dios, Oier Lopez de Lacalle et al.

Recent work has shown that NLP tasks such as Relation Extraction (RE) can be recasted as Textual Entailment tasks using verbalizations, with strong performance in zero-shot and few-shot settings thanks to pre-trained entailment models. The fact that relations in current RE datasets are easily verbalized casts doubts on whether entailment would be effective in more complex tasks. In this work we show that entailment is also effective in Event Argument Extraction (EAE), reducing the need of manual annotation to 50% and 20% in ACE and WikiEvents respectively, while achieving the same performance as with full training. More importantly, we show that recasting EAE as entailment alleviates the dependency on schemas, which has been a road-block for transferring annotations between domains. Thanks to the entailment, the multi-source transfer between ACE and WikiEvents further reduces annotation down to 10% and 5% (respectively) of the full training without transfer. Our analysis shows that the key to good results is the use of several entailment datasets to pre-train the entailment model. Similar to previous approaches, our method requires a small amount of effort for manual verbalization: only less than 15 minutes per event argument type is needed, and comparable results can be achieved with users with different level of expertise.

Iker García-Ferrero, Begoña Altuna, Javier Álvez et al.

Although large language models (LLMs) have apparently acquired a certain level of grammatical knowledge and the ability to make generalizations, they fail to interpret negation, a crucial step in Natural Language Processing. We try to clarify the reasons for the sub-optimal performance of LLMs understanding negation. We introduce a large semi-automatically generated dataset of circa 400,000 descriptive sentences about commonsense knowledge that can be true or false in which negation is present in about 2/3 of the corpus in different forms. We have used our dataset with the largest available open LLMs in a zero-shot approach to grasp their generalization and inference capability and we have also fine-tuned some of the models to assess whether the understanding of negation can be trained. Our findings show that, while LLMs are proficient at classifying affirmative sentences, they struggle with negative sentences and lack a deep understanding of negation, often relying on superficial cues. Although fine-tuning the models on negative sentences improves their performance, the lack of generalization in handling negation is persistent, highlighting the ongoing challenges of LLMs regarding negation understanding and generalization. The dataset and code are publicly available.

Malvina Nissim, Danilo Croce, Viviana Patti et al.

The rapid progress of Large Language Models (LLMs) has transformed natural language processing and broadened its impact across research and society. Yet, systematic evaluation of these models, especially for languages beyond English, remains limited. "Challenging the Abilities of LAnguage Models in ITAlian" (CALAMITA) is a large-scale collaborative benchmarking initiative for Italian, coordinated under the Italian Association for Computational Linguistics. Unlike existing efforts that focus on leaderboards, CALAMITA foregrounds methodology: it federates more than 80 contributors from academia, industry, and the public sector to design, document, and evaluate a diverse collection of tasks, covering linguistic competence, commonsense reasoning, factual consistency, fairness, summarization, translation, and code generation. Through this process, we not only assembled a benchmark of over 20 tasks and almost 100 subtasks, but also established a centralized evaluation pipeline that supports heterogeneous datasets and metrics. We report results for four open-weight LLMs, highlighting systematic strengths and weaknesses across abilities, as well as challenges in task-specific evaluation. Beyond quantitative results, CALAMITA exposes methodological lessons: the necessity of fine-grained, task-representative metrics, the importance of harmonized pipelines, and the benefits and limitations of broad community engagement. CALAMITA is conceived as a rolling benchmark, enabling continuous integration of new tasks and models. This makes it both a resource -- the most comprehensive and diverse benchmark for Italian to date -- and a framework for sustainable, community-driven evaluation. We argue that this combination offers a blueprint for other languages and communities seeking inclusive and rigorous LLM evaluation practices.

Oscar M Cumbicus-Pineda, Iker Gutiérrez-Fandiño, Itziar Gonzalez-Dios et al.

In this paper we present a simple re-ranking method for Automatic Sentence Simplification based on the noisy channel scheme. Instead of directly computing the best simplification given a complex text, the re-ranking method also considers the probability of the simple sentence to produce the complex counterpart, as well as the probability of the simple text itself, according to a language model. Our experiments show that combining these scores outperform the original system in three different English datasets, yielding the best known result in one of them. Adopting the noisy channel scheme opens new ways to infuse additional information into ATS systems, and thus to control important aspects of them, a known limitation of end-to-end neural seq2seq generative models.

Margot Madina, Itziar Gonzalez-Dios, Melanie Siegel

Easy-to-Read Language (E2R) is a controlled language variant that makes any written text more accessible through the use of clear, direct and simple language. It is mainly aimed at people with cognitive or intellectual disabilities, among other target users. Plain Language (PL), on the other hand, is a variant of a given language, which aims to promote the use of simple language to communicate information. German counts with Leichte Sprache (LS), its version of E2R, and Einfache Sprache (ES), its version of PL. In recent years, important developments have been conducted in the field of LS. This paper offers an updated overview of the existing Natural Language Processing (NLP) tools and resources for LS. Besides, it also aims to set out the situation with regard to LS and ES in Germany.

Jaione Bengoetxea, Itziar Gonzalez-Dios, Rodrigo Agerri

Recent research on dialectal NLP has identified data scarcity as a primary limitation. To address this limitation, this paper presents a catalog of contemporary Basque dialectal data and resources, offering a systematic and comprehensive compilation of the dialectal data currently available in Basque. Two types of data sources have been distinguished: online data originally written in some dialect, and standard-to-dialect adapted data. The former includes all dialectal data that can be found online, such as news and radio sites, informal tweets, as well as online resources such as dictionaries, atlases, grammar rules, or videos. The latter consists of data that has been adapted from the standard variety to dialectal varieties, either manually or automatically. Regarding the manual adaptation, the test split of the XNLI Natural Language Inference dataset was manually adapted into three Basque dialects: Western, Central, and Navarrese-Lapurdian, yielding a high-quality parallel gold standard evaluation dataset. With respect to the automatic dialectal adaptation, the automatically adapted physical commonsense dataset (BasPhyCowest) underwent additional manual evaluation by native speakers to assess its quality and determine whether it could serve as a viable substitute for full manual adaptation (i.e., silver data creation).

Jaione Bengoetxea, Itziar Gonzalez-Dios, Rodrigo Agerri

Physical commonsense reasoning represents a fundamental capability of human intelligence, enabling individuals to understand their environment, predict future events, and navigate physical spaces. Recent years have witnessed growing interest in reasoning tasks within Natural Language Processing (NLP). However, no prior research has examined the performance of Large Language Models (LLMs) on non-question-answering (non-QA) physical commonsense reasoning tasks in low-resource languages such as Basque. Taking the Italian GITA as a starting point, this paper addresses this gap by presenting BasPhyCo, the first non-QA physical commonsense reasoning dataset for Basque, available in both standard and dialectal variants. We evaluate model performance across three hierarchical levels of commonsense understanding: (1) distinguishing between plausible and implausible narratives (accuracy), (2) identifying the conflicting element that renders a narrative implausible (consistency), and (3) determining the specific physical state that creates the implausibility (verifiability). These tasks were assessed using multiple multilingual LLMs as well as models pretrained specifically for Italian and Basque. Results indicate that, in terms of verifiability, LLMs exhibit limited physical commonsense capabilities in low-resource languages such as Basque, especially when processing dialectal variants.

Kepa Bengoetxea, Itziar Gonzalez-Dios

Readability assessment is the task of determining how difficult or easy a text is or which level/grade it has. Traditionally, language dependent readability formula have been used, but these formulae take few text characteristics into account. However, Natural Language Processing (NLP) tools that assess the complexity of texts are able to measure more different features and can be adapted to different languages. In this paper, we present the MultiAzterTest tool: (i) an open source NLP tool which analyzes texts on over 125 measures of cohesion,language, and readability for English, Spanish and Basque, but whose architecture is designed to easily adapt other languages; (ii) readability assessment classifiers that improve the performance of Coh-Metrix in English, Coh-Metrix-Esp in Spanish and ErreXail in Basque; iii) a web tool. MultiAzterTest obtains 90.09 % in accuracy when classifying into three reading levels (elementary, intermediate, and advanced) in English and 95.50 % in Basque and 90 % in Spanish when classifying into two reading levels (simple and complex) using a SMO classifier. Using cross-lingual features, MultiAzterTest also obtains competitive results above all in a complex vs simple distinction.

Francesca De Luca Fornaciari, Begoña Altuna, Itziar Gonzalez-Dios et al.

In this work, we explore idiomatic language processing with Large Language Models (LLMs). We introduce the Idiomatic language Test Suite IdioTS, a new dataset of difficult examples specifically designed by language experts to assess the capabilities of LLMs to process figurative language at sentence level. We propose a comprehensive evaluation methodology based on an idiom detection task, where LLMs are prompted with detecting an idiomatic expression in a given English sentence. We present a thorough automatic and manual evaluation of the results and an extensive error analysis.

Jaione Bengoetxea, Itziar Gonzalez-Dios, Rodrigo Agerri

In this paper, we evaluate the capacity of current language technologies to understand Basque and Spanish language varieties. We use Natural Language Inference (NLI) as a pivot task and introduce a novel, manually-curated parallel dataset in Basque and Spanish, along with their respective variants. Our empirical analysis of crosslingual and in-context learning experiments using encoder-only and decoder-based Large Language Models (LLMs) shows a performance drop when handling linguistic variation, especially in Basque. Error analysis suggests that this decline is not due to lexical overlap, but rather to the linguistic variation itself. Further ablation experiments indicate that encoder-only models particularly struggle with Western Basque, which aligns with linguistic theory that identifies peripheral dialects (e.g., Western) as more distant from the standard. All data and code are publicly available.

Xavier Gómez Guinovart, Itziar Gonzalez-Dios, Antoni Oliver et al.

Language resources are necessary for language processing,but building them is costly, involves many researches from different areas and needs constant updating. In this paper, we describe the crosslingual framework used for developing the Multilingual Central Repository (MCR), a multilingual knowledge base that includes wordnets of Basque, Catalan, English, Galician, Portuguese, Spanish and the following ontologies: Base Concepts, Top Ontology, WordNet Domains and Suggested Upper Merged Ontology. We present the story of MCR, its state in 2017 and the developed tools.

Javier Álvez, Itziar Gonzalez-Dios, German Rigau

We describe a detailed analysis of a sample of large benchmark of commonsense reasoning problems that has been automatically obtained from WordNet, SUMO and their mapping. The objective is to provide a better assessment of the quality of both the benchmark and the involved knowledge resources for advanced commonsense reasoning tasks. By means of this analysis, we are able to detect some knowledge misalignments, mapping errors and lack of knowledge and resources. Our final objective is the extraction of some guidelines towards a better exploitation of this commonsense knowledge framework by the improvement of the included resources.

Javier Álvez, Itziar Gonzalez-Dios, German Rigau

Most commonly, the Open World Assumption is adopted as a standard strategy for the design, construction and use of ontologies. This strategy limits the inferencing capabilities of any system because non-asserted statements (missing knowledge) could be assumed to be alternatively true or false. As we will demonstrate, this is especially the case of first-order logic (FOL) ontologies where non-asserted statements is nowadays one of the main obstacles to its practical application in automated commonsense reasoning tasks. In this paper, we investigate the application of the Closed World Assumption (CWA) to enable a better exploitation of FOL ontologies by using state-of-the-art automated theorem provers. To that end, we explore different CWA formulations for the structural knowledge encoded in a FOL translation of the SUMO ontology, discovering that almost 30 % of the structural knowledge is missing. We evaluate these formulations on a practical experimentation using a very large commonsense benchmark obtained from WordNet through its mapping to SUMO. The results show that the competency of the ontology improves more than 50 % when reasoning under the CWA. Thus, applying the CWA automatically to FOL ontologies reduces their ambiguity and more commonsense questions can be answered

Javier Álvez, Itziar Gonzalez-Dios, German Rigau

In this paper, we report on the practical application of a novel approach for validating the knowledge of WordNet using Adimen-SUMO. In particular, this paper focuses on cross-checking the WordNet meronymy relations against the knowledge encoded in Adimen-SUMO. Our validation approach tests a large set of competency questions (CQs), which are derived (semi)-automatically from the knowledge encoded in WordNet, SUMO and their mapping, by applying efficient first-order logic automated theorem provers. Unfortunately, despite of being created manually, these knowledge resources are not free of errors and discrepancies. In consequence, some of the resulting CQs are not plausible according to the knowledge included in Adimen-SUMO. Thus, first we focus on (semi)-automatically improving the alignment between these knowledge resources, and second, we perform a minimal set of corrections in the ontology. Our aim is to minimize the manual effort required for an extensive validation process. We report on the strategies followed, the changes made, the effort needed and its impact when validating the WordNet meronymy relations using improved versions of the mapping and the ontology. Based on the new results, we discuss the implications of the appropriate corrections and the need of future enhancements.