LANLP focuses on community-driven resource development and evaluation for Iberian languages, and diverse Latin American languages (including indigenous and minority languages). We aim to bridge regional communities to share initiatives, corpora and tools. LANLP fills this gap, fostering new contacts between Iberian and Latin American NLP research groups. The goals are to (1) highlight challenges in processing these languages, (2) share novel datasets and models, and (3) catalyze future collaborations and shared tasks. We emphasize both academic rigor and community inclusivity, encouraging contributions from established researchers and grassroots language advocates alike.

Artificial intelligence systems are often presented as universal, yet their interaction paradigms remain predominantly text-first, limiting alignment with primarily oral languages and communicative practices. Using Guaraní, an official and widely spoken language of Paraguay, as a motivating case, this work examines how language support risks remaining symbolic when spoken interaction is reduced to a speech-to-text interface. We explore an oral-first, multi-agent framing in which turn-taking, repair, shared context, and governance are treated as core components of interaction rather than peripheral features. By separating language understanding from the conversation state and permission mechanisms, the architecture makes conversational structure and control explicit, enabling reasoning over interaction dynamics rather than isolated commands. Framing conversational coordination as a cognitively motivated reasoning problem over shared state connects insights from human dialogue to the design of AI systems that are more interpretable and responsive in oral and low-resource settings.

In this paper, we present AI-TraLow, a project dedicated to advancing AI-driven translation for low-resource languages and cultures. The research is structured around three primary objectives: firstly, the development of advanced data curation techniques designed to refine parallel corpora and detect machine-generated content; secondly, the exploration of integrating structured linguistic resources—such as dictionaries and grammatical rules—directly into model prompts and fine-tuning techniques to enhance translation precision; and thirdly, the mitigation of hardware constraints through knowledge distillation to produce efficient models viable for standard desktop environments. By targeting specific linguistic groups, including Iberian varieties (Aranese, Aragonese, and Asturian), Mayan languages, and languages of vulnerable migrant communities, AI-TraLow seeks to foster linguistic diversity and digital inclusion. Ultimately, this initiative delivers open-source tools and models that ensure cultural heritage is both preserved and accessible within the contemporary digital landscape.

The availability of open resources and corpora is a fundamental requirement for research in Natural Language Processing (NLP) and Computational Linguistics; however, languages spoken in Latin America and the Iberian Peninsula, particularly Indigenous, minority, and regional varieties, remain structurally under-resourced and under-represented. This paper presents a historical account of OpenCor (Latin American and Iberian Languages Open Corpora Forum), a community-driven initiative created to promote, document, and discuss open linguistic corpora and lexical resources for these languages. Conceived as a collaborative forum rather than a competitive evaluation venue, OpenCor focuses on data creation, licensing practices, sustainability, and community building. Between 2018 and 2024, OpenCor was organized as a recurring workshop co-located with major conferences, fostering dialogue across countries, institutions, and linguistic traditions. By documenting the initiative’s motivations, organizational trajectory, submission trends, and the diversity of resources presented, this paper aims to preserve institutional memory, highlight the often-invisible labor of corpus development, and provide a reference for future initiatives dedicated to openness and linguistic diversity.

Although healthcare documentation is increasingly dependent on speech-based clinical interactions, general-purpose Automatic Speech Recognition (ASR) and Large Language Models (LLMs) lack the domain adaptation, structured control and interoperability guarantees required in regulated medical environments. These limitations often result in transcription errors, hallucinated content, and limited alignment with standardized coding systems. This paper introduces MedicaLLM, a multilingual, end-to-end framework integrating domain-adapted ASR, LLM-based structured report generation, and ontology-driven semantic enrichment within a modular architecture for clinical documentation. MedicaLLM combines medical interview transcription with structured report generation, summarization, and error correction; Named Entity Recognition (NER); and Medical Entity Linking (MEL) to align with standards such as SNOMED-CT and ICD-10. Deployed as a secure software as a service (SaaS) platform with REST API integration, MedicaLLM aims to reduce the administrative burden, improve the quality of documentation, and enhance semantic interoperability across healthcare systems, all while maintaining computational efficiency and clinical reliability.

Customer service and incident management increasingly rely on multimodal evidence, combining text, images and audio. However, general-purpose models lack domain grounding, structured output control and reliability guarantees required in regulated enterprise environments, often leading to hallucinated responses and limiting their practical deployment. This paper presents mCS-LM, a multilingual multimodal framework that integrates Large Language Models (LLMs), Visual Language Models (VLMs), Audio Language Models (ALMs) and Retrieval-Augmented Generation (RAG) within a modular and traceable architecture tailored to customer service and incident management. The system introduces complementary processing flows: (i) perception modules for visual and audio understanding aligned with LLM-based reasoning, and (ii) structured report generation from multimodal evidence through supervised fine-tuning using QLoRA and efficient adaptation techniques. To mitigate hallucinations and improve factual reliability, the framework incorporates vector databases and multimodal RAG pipelines that retrieve domain-specific knowledge from external corporate sources. Formal structural schemas and validation mechanisms enforce output consistency and syntactic correctness. The platform is deployed as a web-based system with REST API integration, enabling scalable multimodal interaction across channels such as instant messaging, email and web chat. Experimental results demonstrate that multimodal generative models can be specialized for structured, domain-constrained enterprise tasks while maintaining computational viability and robustness.

Los Grandes Modelos de Lenguaje (LLMs) abren oportunidades para el Procesamiento del Lenguaje Natural (PLN) en contextos institucionales, si bien plantean riesgos críticos en entornos regulados y multilingües, especialmente en lo relativo a protección de datos personales, trazabilidad de decisiones y equidad entre lenguas con distinta disponibilidad de recursos. Presentamos SAFEWORDs, proyecto que acaba de iniciarse en el marco del proyecto coordinado “HumanAIze” (Plan Nacional de Inteligencia Artificial 2025, España), que propone un marco reproducible de privacy-by-design y ethics-by-design para la evaluación y alineación de LLMs en las lenguas oficiales de la Península Ibérica (español, catalán, valenciano, gallego y euskera). El marco integra: (i) anonimización automática conforme al RGPD, con protocolos explícitos de detección de fuga residual y verificación adversarial; (ii) transformación orientada a la accesibilidad textual y al lenguaje claro; y (iii) evaluación en el dominio biomédico, donde la sensibilidad de los datos y la precisión terminológica exigen mecanismos adicionales de control generativo. Desde el punto de vista metodológico, se comparan configuraciones zero-shot y few-shot, y se documentan prompts, hiperparámetros y recursos para facilitar la replicabilidad y la gobernanza de recursos. Además de sintetizar resultados de referencia de la literatura para contextualizar métricas y órdenes de magnitud esperables, el trabajo discute implicaciones éticas y limitaciones del enfoque propuesto. La propuesta se alinea con las líneas de trabajo de SEPLN y con los objetivos de LANLP, al establecer protocolos transferibles para el desarrollo de tecnologías lingüísticas confiables en ecosistemas caracterizados por variación dialectal y lenguas infrarepresentadas.

Transformer-based language models, ubiquitous in NLP nowadays, generate internal representations (embeddings) of words and sentences. Yet, systematic comparisons of embedding strategies from various models remain limited. In this work, we evaluate Spanish embeddings from several BERT-like models (BETO, multilingual BERT, XLM-RoBERTa, ROUBERTa) to understand their syntactic and semantic capabilities across layers. We propose novel sentence-level analogy tests to probe generalization. Results show tasks like verb negation or word reordering perform best with embeddings from earlier layers, while nuanced semantic distinctions—such as agent or patient gender—are better captured by deeper layers. Our findings provide guidelines for embedding strategies and offer a foundation for further NLP research.