Donostia International Physics Center
Archivo Digital para la Docencia y la InvestigaciónNot a member yet
46642 research outputs found
Sort by
Cross-lingual Transfer for Low-Resource Natural Language Processing
No full text232 p.Natural Language Processing (NLP) has seen remarkable advances in recent years, particularly with the emergence of Large Language Models that have achieved unprecedented performance across many tasks. However, these developments have mainly benefited a small number of high-resource languages such as English. The majority of languages still face significant challenges due to the scarcity of training data and computational resources. To address this issue, this thesis focuses on cross-lingual transfer learning, a research area aimed at leveraging data and models from high-resource languages to improve NLP performance for low-resource languages. Specifically, we focus on Sequence Labeling tasks such as Named Entity Recognition, Opinion Target Extraction, and Argument Mining. The research is structured around three main objectives: (1) advancing data-based cross-lingual transfer learning methods through improved translation and annotation projection techniques, (2) developing enhanced model-based transfer learning approaches utilizing state-of-the-art multilingual models, and (3) applying these methods to real-world problems while creating open-source resources that facilitate future research in low-resource NLP. More specifically, this thesis presents a new method to improve data-based transfer with T-Projection, a state-of-the-art annotation projection method that leverages text-to-text multilingual models and machine translation systems. T-Projection significantly outperforms previous annotation projection methods by a wide margin. For model-based transfer, we introduce a constrained decoding algorithm that enhances cross-lingual Sequence Labeling in zero-shot settings using text-to-text models. Finally, we develop Medical mT5, the first multilingual text-to-text medical model, demonstrating the practical impact of our research on real-world applications
Development of a tool for thermoeconomic diagnosis of thermal facilities in buildings by applying artificial intelligence
No full text229 p.Buildings are responsable for 40% of total energy consumption in the European Union and 36% of energy-realted greenhouse gas emissions. Heating, ventilation, and air conditioning systems account for nearly half of the energy used in buildings and approximately 10-20% of overall energy demand. These high consumption levels, coupled with rising greenhouse gas emissions, contribute to global temperature increases, which have negative consequences for both the environment and society. To address this issue, the Energy Union was established, setting intermediate targets for 2030 and long-term objectives for 2050. One of its five key pillars is energy efficiency. Given that thermal system efficiency can decline by approximately 30% over a 20-year period, ensuring effective maintenance through preventive strategies is crucial for optimising energy use and reducing environmental impact.In this context, Chapter 1 provides and overview of the current state of energy efficiency and sustainability in buildings, as well as the role of advanced fault detection and diagnosis techniques. The chapter highlights the advantages of hybrid approaches, particularly the integration of thermoeconomics and Artificial Intelligence ¿ two fields that are increasingly being combined to enhance system analysis. Specifically, it explores two advanced thermoeconomic diagnosis methodologies: characteristic curves and exergetic operators. Additionally, within the realm of Artificial Intelligence, special attention is given to Machine Learning, with a focus on artificial neural networks.Building on this foundation, Chapter 2 presents the primary objective of this thesis: the development of a tool capable of identifying faults in building thermal systems and quantifying their impact. By combining thermoeconomic diagnostics with Machine Learning, this tool aims to minimise the negative effects of inefficiencies in thermal systems.One of the fundamental challenges in analysing building thermal systems is their dynamic nature. These systems constantly adapt to fluctuating user demands and changing external conditions. Therefore, it is essential to study them from a dynamic perspective. To address this, Chapter 3 examines two case studies involving typical building thermal facilities using the only existing thermoeconomic software capable of performing dynamic analyses.Expanding upon this research, Chapter 4 takes a step further by optimising the thermoeconomic software to make it freely available as an open-source tool. This enhancement not only ensures accessibility for a wider audience but also eliminates the need for manual calculations, significantly improving usability¿particularly for maintenance personnel managing building thermal systems. The newly developed tool has been applied to a case study involving two typical building thermal systems in different locations, demonstrating its practical applicability. The tool's code is provided in the Appendix.Finally, Chapter 5 introduces an innovative methodology for real-time fault detection and impact quantification. This methodology integrates the software developed in Chapter 4 and leverages Machine Learning to simulate the ideal operating behaviour of thermal systems at any given moment. By continuously comparing the real-time performance of the system to its ideal reference model, this approach enables accurate thermoeconomic diagnosis, improving energy efficiency and system reliabilit
Unravelling the molecular mechanisms of unexplained male infertility
No full text325 p.Lifetime infertility affects approximately 17.5 % of couples worldwide. Although assisted reproductive technology (ART) treatments have offered an effective treatment for infertility, success rates are still far from optimal. In fact, only 30 % of transferred embryos achieve an ongoing pregnancy. While the focus of infertility has historically been on the female factor, male factor infertility accounts for nearly 50 % of cases. Semen analysis is the cornerstone of evaluating male infertility, but it remains an imperfect and often insufficient diagnostic tool. As a result, around 30 % of the men are infertile despite having normal semen parameters, medical history, and physical examinations. This condition is known as unexplained male infertility (UMI). The lack of classification criteria and diagnostic tools makes UMI a diagnosis by exclusion that comprises a heterogeneous group of patients that often go unnoticed in routine semen analysis. As a result, UMI has become a challenge for modern andrology, indicating that factors beyond traditional semen analysis, such as molecular, genetic, or functional abnormalities in sperm, may contribute to infertility in these cases. Consequently, understanding UMI condition remains critical to optimizing ART treatments. In this context, this study aims to characterize the molecular mechanisms underlying UMI to contribute to the understanding of male infertility and improve diagnostic and treatment options. By transcriptomic and proteomic approaches, we identified significant alterations in energy metabolism associated processes in capacitated sperm from UMI patients compared to healthy donors, mainly in mitochondrial oxidative phosphorylation (OXPHOS). Functional assays validated these findings, revealing increased NADH and succinate dehydrogenase activity, which confirm disruptions in energy metabolism and redox balance in UMI patients. Furthermore, ART outcomes from the patients analyzed in this study demonstrated reduced ongoing pregnancy rates compared to donors, underscoring the clinical relevance of these metabolic disruptions. While our analysis of pooled UMI samples provided valuable insights into the potential mechanisms underlying UMI pathology, pooling samples may mask inter-individual variability in molecular and clinical profiles. To address this, we employed individualized transcriptomic approaches to validate findings from the pooled analyses. Our results confirmed that individual UMI patients exhibit mitochondrial-encoded transcriptome alterations in capacitated sperm, showing a downregulation of OXPHOS coding genes and an upregulation of non-coding genes. These findings suggest that mitochondrial dysfunction could be involved in UMI pathogenesis and in the impairment of fertilizing ability of the male gamete in UMI patients. To the best of our knowledge, this is the first study to analyze sperm transcriptomes from individual UMI patients, effectively overcoming the challenges posed by patient-to-patient heterogeneity. Importantly, this individualized approach enabled us to categorize UMI patients into two subgroups: donor-like and UMI-like, based on their transcriptomic similarity to sperm donor pools or to UMI patient pools with poor ART outcomes. Furthermore, we identified MT-RNR2 and RN7SKP203 as potential biomarkers for distinguishing UMI-like patients. In addition to transcriptomic insights, we explored other potential contributors to UMI. Characterization of anti-sperm antibody (ASA) patterns in seminal plasma revealed that research in the role of immune system in the male reproductive tract could also help uncover underlying causes of UMI
Basque and Occitan in the new global context. Some reflections on language coexistence in multilingual education
No full textThis chapter examines the coexistence of minority, majority, and global languages within multilingual education, focusing on Basque and Occitan in two distinct European contexts: New Aquitaine (France) and the Basque Autonomous Community (Spain). Through a comparative lens, it is analyzed how educational policies shape the status and integration of these languages. In New Aquitaine, Occitan and Basque remain optional subjects alongside compulsory French and English, reflecting a model where minority languages occupy a peripheral role. Conversely, the Basque Autonomous Community enforces Basque as a compulsory language, positioning it alongside Spanish and English in a reinforced trilingual curriculum. The discussion situates these regional approaches within broader global trends, emphasizing tensions between local linguistic revitalization efforts and the pervasive influence of English as a global lingua franca. A paradigm shift is proposed toward genuinely integrated multilingual frameworks that transcend additive models, advocating for curricular designs that foster balanced coexistence and sustainability of linguistic diversity in an increasingly globalized educational landscape
Research/action laboratory for education in social sustainability from graphic activism
No full text[EN] Campus Bizia Lab is an educational program from the University of the Basque
Country (UPV/EHU) based on the IKD i3 model, which integrates learning (Ikaskuntza),
research (Ikerketa), and sustainability (Iraunkortasuna). The program aims to address social
sustainability challenges in the daily life of students within the UPV/EHU campus.
Methodology: The projects within Campus Bizia Lab are driven by active, transdisciplinary,
cooperative, and research-based teaching and learning strategies. These projects are realized
through Bachelor's Final Projects (TFG) that address the challenge of “Education for Social
Transformation - Equality in Diversity”. Results: The TFGs developed in the Degrees of
Creation and Design, and Audiovisual Communication and Advertising and Public Relations
at UPV/EHU focus on the role of graphic design to generate emotional functions that help
create new structures of understanding and overcome barriers to collective transformation.
Discussion and Conclusions: These projects demonstrate how graphic design can play a key
role in social transformation, contributing to an approach of equality in diversity and
promoting positive change within the educational community.Campus Bizia Lab es un programa educativo de la Universidad del País Vasco
(UPV/EHU) basado en el modelo IKD i3, que integra el aprendizaje (Ikaskuntza), la
investigación (Ikerketa) y la sostenibilidad (Iraunkortasuna). El objetivo del programa es
resolver retos de sostenibilidad social en la vida cotidiana del alumnado dentro del propio
campus de la UPV/EHU. Metodología: Los proyectos dentro de Campus Bizia Lab se
impulsan mediante estrategias de enseñanza-aprendizaje activas, transdisciplinares,
cooperativas y basadas en la investigación. Estos proyectos se materializan a través de Trabajos
Finales de Grado (TFG) que abordan el reto de “Educación para la Transformación Social -
Igualdad en la diversidad”. Resultados: Los TFG desarrollados en los Grados de Creación y
Diseño, y Comunicación Audiovisual y Publicidad y Relaciones Públicas, de la UPV/EHU, se
enfocan en el protagonismo del diseño gráfico para generar funciones emocionales que ayuden
a crear nuevas estructuras de comprensión y superen barreras hacia la transformación colectiva. Discusión y Conclusiones: Estos proyectos evidencian cómo el diseño gráfico puede
jugar un papel fundamental en la transformación social, contribuyendo a un enfoque de
igualdad en la diversidad y promoviendo un cambio positivo dentro de la comunidad
educativa.Proyecto Campus Bizia Lab con título «Igualdad en la diversidad. Activismo gráfico en la UPV/EHU para la transformación social» subvencionado por la Dirección de Sostenibilidad y Compromiso Social del Vicerrectorado de Desarrollo Científico-Social y Transferencia de la UPV/EHU
Disruptive Climate Protests Against Companies: Analyzing Customer Perceptions in an Experimental Setting
No full textIn response to insufficient progress on carbon neutrality commitments, climate activism is
increasingly targeting companies—both oil companies and the banks that finance their activity.
Drawing on Attribution Theory, this research investigates how protest extremity (moderate vs.
disruptive) and target type (oil companies vs. banks funding oil projects) influence perceived
corporate agency and consumer behavior. In Study 1, a moderated mediation model revealed
that moderate protests were more effective than disruptive ones, particularly when aimed at
banks. Disruptive protests reduce perceived corporate agency and responsibility, thereby
lowering consumers’ intentions to switch providers. In Study 2, we manipulated the bank’s role
in exacerbating climate change (direct vs. indirect), but this framing did not significantly affect
customer perceptions—suggesting that protest extremity, rather than target type, is the primary
driver lowering perceived agency and switching intentions. Implications for climate activism
strategies and corporate reputation management are discussed.This work was supported by the Eusko Jaurlaritza (IT1731-22); the Ministerio de Ciencia, Innovación y Universidades (AEI/10.13039/501100011033); and the European Regional Development Fund (PID2021-123686OB-I00)
Characterisation of PGC1α-driven IFN. Activation: Instrinsic and extrinsic effects and crosstalk with cancer-associated fibroblasts in prostate cancer
No full text326 p.The transcriptional co-regulator peroxisome proliferator-activated receptor gamma co-activator 1 alpha (PGC1α) has been identified as a tumour and metastasis suppressor in prostate cancer, regulating genes involved in the cell cycle, extracellular matrix and interferon pathway. In PC3 cells, PGC1α activates the interferon cascade both intrinsically and extrinsically, in cooperation with its partner oestrogen-related receptor alpha (ERRα). This work provides an in-depth characterisation of the molecular mechanisms driving interferon pathway activation by PGC1α. The canonical interferon signalling and its transcriptional program were confirmed to be induced. Although the main trigger remains unidentified, mitochondrial DNA contributes to sustained activation. PGC1α also modulates the secretion of pro-inflammatory cytokines and genes involved in antigen presentation, suggesting communication with immune cells like T lymphocytes. Paracrine effects on normal and cancer-associated fibroblasts were also explored, though no similar phenotype was observed as in epithelial cells. These findings offer new insights into how PGC1α-expressing prostate cancer cells interact with the tumour microenvironment. However, more complex co-culture assays are needed to further explore these mechanismsLa realización de esta tesis doctoral ha sido posible gracias a la concesión del contrato de formación de personal investigador en la UPV/EHU, convocatoria 2019, de la Universidad del País Vasco/ Euskal Herriko Unibertsitatea (UPV/EHU) (Contrato PIF 19/81).
La financiación para la realización de la estancia internacional en el laboratorio de la Doctora Sara Zanivan en el Cancer Research UK Scotland Institute ha sido proporcionada por European Molecular Biology Organization (EMBO) (EMBO Scientific Exchange Grant, SEG Number: 10688)
El espacio educativo como experiencia de aprendizaje: una aproximación al paradigma de la arquitectura pedagógica
No full text907 p.Esta tesis doctoral parte desde un paradigma interpretativo que concibe el espacio educativo como una experiencia de aprendizaje y un factor pedagógico esencial en el desarrollo integral del alumnado, situándose en la intersección entre arquitectura y educación y en constante diálogo con múltiples ámbitos como la sociología, la neurociencia, la pedagogía y/o la psicología entre otros. Históricamente, la arquitectura escolar se ha estudiado desde la perspectiva patrimonial y como reflejo de la evolución del sistema educativo y la sociedad en cada época. Su desarrollo ha estado condicionado por corrientes higienistas y funcionalistas, priorizando criterios técnicos y económicos por encima de consideraciones pedagógicas. Sin embargo, desde el paradigma de la educación del siglo XXI, los retos de la arquitectura educativa son bien distintos: la incertidumbre, la complejidad, la inmediatez, la intersección entre lo tangible y lo intangible, lo virtual y lo real, lo artificial y lo humano, etc. ¿Cómo diseñar un entorno educativo que responda a lo cambiante y a las necesidades emergentes del aprendizaje?La hipótesis central de la presente tesis doctoral plantea que el espacio educativo, entendido como un entorno dinámico con múltiples características sensoriales, influye en las emociones, las relaciones y el aprendizaje del alumnado. Desde esta perspectiva, la arquitectura escolar no debe limitarse al control y la contención del alumnado, sino que debe ser una invitación al aprendizaje a través de la exploración, la experimentación, la introspección, la interacción y la socialización. Dicho de otra forma, el espacio educativo debe concebirse como un escenario que propicie experiencias significativas y autónomas, enriqueciendo la vivencia de la escuela
Neuroplasticity in patients with brain tumors: insights from structural and functional perspective
No full text130 p.Esta tesis investiga cómo los tumores cerebrales inducen cambios neuroplásticos en el cerebro. Utilizando técnicas avanzadas de neuroimagen como MRI y fMRI, se exploran cambios estructurales y funcionales. Para ello, se comparan pacientes con tumores y adultos sanos, evaluando su progresión antes y después de la intervención quirúrgica para la resección del tumor. Los resultados muestran que los tumores provocan adaptaciones que van más allá del área afectada, extendiéndose por todo el cerebro, lo que resalta la naturaleza interconectada del cerebro. Estructuralmente, el volumen de materia gris cambia antes y después de la operación, permitiendo diferenciar entre pacientes y personas sanas. A nivel funcional, se observan cambios en el patrón de conectividad en ambos hemisferios antes de la cirugía, los cuales se mantienen después de la operación, demostrando la efectividad tanto de los mecanismos de neuroplasticidad como del enfoque quirúrgico. Los resultados de este trabajo doctoral destacan la importancia de considerar los efectos globales de los tumores cerebrales y el uso de enfoques basados en redes para entender la neuroplasticidad y desarrollar estrategias terapéuticas más efectivas.The research presented in this thesis was partially supported by grants RTI2018 093547 B I00 (LANGCONN) by the Spanish Ministry of Economy and Competitiveness; by the Spanish Ministry of Science, Innovation and Universities through the predoctoral grant PRE2019-091492 awarded to Lucía Manso Ortega; and Ramon y Cajal Fellowship RYC2022-035533-I awarded to Ileana Quiñone
Deportados vascos en los KL de la Alemania nazi (1940-1945)
No full text888 p.La tesis analiza la deportación de 262 vascos de Hegoalde hacia los campos de concentración del Reichalemán en el periodo 1940-1945. Para comprender cómo se gestó este fenómeno la investigación seretrotrae hasta el golpe de Estado que el 18 de julio de 1936 se dio en la Segunda República española,momento en el que surgió el exilio vasco. Continúa con un examen de los mecanismos represivos quepadeció el exilio vasco en la Europa ocupada, especialmente en la República francesa, entre 1939 y 1944.Estos mecanismos originaron que la deportación vasca hacia la Alemania nazi se desarrollara en dos fases: entre 1940-1941 la primera y entre 1942-1945 la segunda. Una vez en los KL, los vascos fueron víctimas del sistema concentracionario nazi, sistema represivo sin parangón y no equiparable a otros contemporáneos en los que también hubo presos vascos. Finalmente, la investigación ahonda en e lhecho de que la deportación no terminó con la liberación de los supervivientes en 1945, ya que las muertes prematuras y las enfermedades crónicas se prolongaron el resto de sus vidas