43569 research outputs found
Sort by
El objeto como mediación: plan pedagógico para el museo frente a discursos antiinmigración
No full textEste artículo presenta una propuesta pedagógica dirigida a institutos de secundaria,
desarrollada en el barrio pesquero del Molinar de Palma para el Museo Marítimo de Mallorca (MMM). Su objetivo es fomentar la comunicación intergeneracional e intercultural y contrarrestar los discursos antiinmigración. La iniciativa forma parte de una investigación más amplia impulsada por el museo durante 2023 y 2024. Desde la perspectiva de la Antropología de la comunicación, se explora cómo los objetos pueden resignificarse como artefactos, en tanto que constructos simbólicos. El objeto devenido artefacto impugna las lógicas temporales aceleracionistas, la descontextualización y la homogeneización, y apela a la articulación comunitaria. La propuesta se articula en torno a 20 objetos cargados de sentido, representativos de la memoria de la comunidad pesquera del barrio. Al ser interpretados como elementos de mediación, estos objetos se transforman en artefactos que facilitan la construcción de significados compartidos. De ellos, diez son utilizados como catalizadores de preguntas destinadas a cuestionar el marco socialmente aceptado sobre la migración. En la discusión se analiza cómo el proyecto posiciona al museo como un espacio de diálogo y encuentro, desde el cual intervenir —gracias a su poder discursivo— en uno de los grandes desafíos del siglo XXI: transformar las estructuras coloniales del discurso que alimentan los relatos de odio y exclusión, y hacer frente al fundamentalismo cultural que ha sostenido los discursos hegemónicos
Comprehensive genomic characterization of early-stage bladder cancer
No full textUnderstanding the molecular landscape of nonmuscle-invasive bladder cancer (NMIBC) is essential to improve risk assessment and treatment regimens. We performed a comprehensive genomic analysis of patients with NMIBC using whole-exome sequencing (n = 438), shallow whole-genome sequencing (n = 362) and total RNA sequencing (n = 414). A large genomic variation within NMIBC was observed and correlated with different molecular subtypes. Frequent loss of heterozygosity in FGFR3 and 17p (affecting TP53) was found in tumors with mutations in FGFR3 and TP53, respectively. Whole-genome doubling (WGD) was observed in 15% of the tumors and was associated with worse outcomes. Tumors with WGD were genomically unstable, with alterations in cell-cycle-related genes and an altered immune composition. Finally, integrative clustering of multi-omics data highlighted the important role of genomic instability and immune cell exhaustion in disease aggressiveness. These findings advance our understanding of genomic differences associated with disease aggressiveness in NMIBC and may ultimately improve patient stratification
Attributing heat-related mortality to climate change
No full textClimate change significantly exacerbates public health risks, particularly heat-related mortality. This PhD thesis examines the impact of human-induced climate change on heat-related deaths in Europe, focusing on regional and demographic disparities. Using Eurostat’s weekly mortality data, stratified by sex and age, the research ensures transparency and comparability. Study I assesses heat-related mortality in 35 European countries, finding that 56% of over 38,000 deaths during the record-breaking summer of 2022 were attributable to climate change. Study II uses extreme value theory to analyze the probability of such events, revealing a tripling likelihood due to climate change. This thesis underscores the pressing need for integrated adaptation and mitigation strategies, emphasizing tailored, region-specific approaches to protect vulnerable populations. It highlights the role of health impact climate attribution in driving climate litigation and shaping policies, contributing valuable insights to the intersection of climate change and public health.El cambio climático agrava significativamente los riesgos para la salud pública, especialmente la mortalidad relacionada con el calor. Esta tesis examina el impacto del cambio climático antropogénico en las muertes relacionadas con el calor en Europa, con un enfoque en las disparidades regionales y demográficas. Utilizando datos semanales de mortalidad de Eurostat, estratificados por sexo y edad, la investigación garantiza transparencia y comparabilidad. El Estudio I analiza la mortalidad por calor en 35 países europeos, encontrando que el 56% de más de 38,000 muertes durante el verano récord de 2022 se atribuyen al cambio climático. El Estudio II utiliza teoría de valores extremos para analizar la probabilidad de tales eventos, revelando una triplicación de su frecuencia debido al cambio climático. Esta tesis resalta la necesidad urgente de estrategias integradas de adaptación y mitigación, enfatizando enfoques específicos para proteger a las poblaciones vulnerables.Programa de Doctorat en Biomedicin
Domain adaptation method and modality gap impact in audio text models for prototypical sound classification
No full textAudio-text models are widely used in zero-shot environmental sound classification as they alleviate the need for annotated data. However, we show that their performance severely drops in the presence of background sound sources. Our analysis reveals that this degradation is primarily driven by SNR levels of background soundscapes, and independent of background type. To address this, we propose a novel method that quantifies and integrates the contribution of background sources into the classification process, improving performance without requiring model retraining. Our domain adaptation technique enhances accuracy across various backgrounds and SNR conditions. Moreover, we analyze the modality gap between audio and text embeddings, showing that narrowing this gap improves classification performance. The method generalizes effectively across state-of-the-art prototypical approaches, showcasing its scalability and robustness for diverse environments
Spatial data science approaches to predict environmental and socioeconomic health determinants
No full textAquesta tesi, organitzada en tres articles, proposa avenços metodològics en els processos de predicció espacial utilitzats per a exposicions ambientals i socioeconòmiques. El primer estudi prediu la temperatura i quatre contaminants atmosfèrics a alta resolució espaciotemporal a Catalunya. Inclou innovacions com l’ús de dades de TROPOMI, l’estimació de la incertesa en les prediccions i mètodes d’explicabilitat per validar les associacions. El segon estudi investiga la inclusió de proxies espacials, com ara coordenades i distàncies, com a predictors en models random forest. En una simulació i dos casos pràctics, concloem que aquestes variables no són sempre beneficioses i identifiquem models alternatius. El tercer estudi prediu la posició socioeconòmica de les llars en zones semirurals de Moçambic, utilitzant un conjunt de dades multimodal format per imatges per satèl·lit i de les llars, i combinant mètodes d’aprenentatge profund i automàtic seguit d’anàlisis d’explicabilitat. Aquesta tesi contribueix a la literatura suggerint metodologies per obtenir estimacions d’exposició més precises, transparents, eficients i informatives.Esta tesis, organizada en tres artículos, propone avances metodológicos en los procesos de predicción espacial utilizados para exposiciones ambientales y socioeconómicas. El primer estudio predice la temperatura y cuatro contaminantes atmosféricos con alta resolución espaciotemporal en Cataluña. Incluye innovaciones como el uso de datos de TROPOMI, la estimación de la incertidumbre en las predicciones y métodos de explicabilidad para validar las asociaciones. El segundo estudio investiga la inclusión de proxies espaciales, como coordenadas y distancias, como predictores en modelos random forest. Mediante una simulación y dos casos prácticos, concluimos que estas variables no siempre son beneficiosas e identificamos modelos alternativos. El tercer estudio predice la posición socioeconómica de los hogares en zonas semirrurales de Mozambique, utilizando un conjunto de datos multimodal formado por imágenes satelitales y de los hogares, y combinando métodos de aprendizaje profundo y automático seguidos de análisis de explicabilidad. Esta tesis contribuye a la literatura sugiriendo metodologías para obtener estimaciones de exposición más precisas, transparentes, eficientes e informativas.Spatial prediction workflows are key for assessing health determinants in epidemiology. This thesis, comprising three manuscripts, advances spatial prediction pipelines for environmental and socioeconomic exposures. The first study predicted temperature and four air pollutants in Catalonia using a multi-exposure framework. Innovations include the use of TROPOMI data, a remote sensing gap-filling evaluation, the estimation of prediction uncertainty, and explainable machine learning to verify associations. The second study investigated spatial proxies (e.g., coordinates and distances) in random forest models. Through simulations and case studies, we found that spatial proxies are not always beneficial, providing insights into their suitability and alternatives. The third study predicted household socioeconomic measures in semi-rural Mozambique using satellite and household imagery. A deep and machine learning pipeline, combined with explainability analyses, revealed clear socioeconomic patterns. This thesis enhances spatial predictions by proposing methodologies for more accurate, transparent, and informative exposure estimates.Universitat Pompeu Fabra. Doctorat en Biomedicin
Mujeres de Libia: una realidad desconocida
No full textDada la escasa información existente sobre el colectivo femenino libio, la presente tesis
visibiliza el caso de las mujeres libias, analiza sus contribuciones e interpreta los cambios
sociales recuperando sus actividades, poco evidenciadas, a causa de las investigaciones
desiguales que a menudo se han realizado sobre las actividades femeninas. Dentro del
marco colonial y poscolonial ocurrido a lo largo del siglo XX, se realiza un análisis crítico
sobre la evolución femenina y el impacto que experimentaron a causa de los múltiples
fenómenos históricos más relevantes: la Colonización Italiana, la Monarquía y el Estado de
Masas/Jamahiriya. Partiendo de la hipótesis de que a pesar de las estructuras patriarcales
propias del colonialismo y de las tradiciones locales y religiosas que han “ocultado” su papel
e incluso han distorsionado su imagen, las mujeres han logrado establecer unas “alianzas”
a través de las reuniones sociales, creando una “sororidad1”, la cual les ha ayudado a
avanzar, tanto en tiempos de guerra, como de paz. Estas “alianzas” han logrado unirlas
hasta el punto de actuar como un movimiento femenino, siéndoles muy útil durante el
sometimiento colonial para resistir la dominación, y en la era poscolonial para transformar
las tradiciones y las costumbres patriarcales intrínsecas en su sociedad.Due to the limited information available on the Libyan female collective, this research makes
visible the case of Libyan women, analyzes their contributions, and interprets the social
changes, recovering their activities, little evidenced, due to the unequal research that has
often been carried out on women's activities. Within the colonial and postcolonial framework
that occurred throughout the 20th century, a critical analysis is carried out on female
evolution and the impact they experienced due to the multiple and most relevant historical
phenomena: the Italian Colonization, the Monarchy, and the Mass State/Jamahiriya.
Starting from the hypothesis that despite the patriarchal structures, typical of colonialism
and local and religious traditions that have “hidden” their role and even distorted their image,
women have managed to establish “alliances” through social meetings, creating a
“sorority2 ”, which has helped them advance, both in times of war and peace. These
“alliances” have managed to unite them to the point of acting as a women's movement, being
very useful to them during colonial subjugation to resist domination, and in the post-colonial
era to transform the patriarchal traditions and customs intrinsic in their society.Programa de Doctorat en Humanitat
Unveiling the role of Her9/Notch pathway in the maintenance of stemness in the hindbrain boundaries
No full textThe generation of cell diversity in the embryonic brain relies on the balance
between cell proliferation and differentiation. The hindbrain boundary cells
specify at the interface between adjacent compartments. Boundary cells display
specialized functions and a specific gene profile. When adjacent compartments
are engaged in neurogenesis, boundary cells remain non-neurogenic. To unveil
how boundary cells maintain their stemness, we combined functional approaches
with high-resolution in vivo imaging in zebrafish embryos. At early stages,
boundary cells display enriched her9 and no Notch activity. Later, boundary cells
lose her9 enrichment and activate Notch3 that triggers the transition from
neuroepithelial to radial glia progenitors that undergo asymmetric divisions. Her9
keeps boundary cells as progenitors by promoting symmetric proliferative
divisions and preventing asymmetric cell divisions. Moreover, Her9 potentially
controls boundary cell proliferation by inhibiting cdkn1ca and promoting
YAP/TAZ-TEAD activity. Altogether, Her9 maintains hindbrain boundary cells as
a progenitor pool at early embryonic stages.La generació de diversitat cel·lular al cervell embrionari recau en el balanç entre
la proliferació i la diferenciació cel·lular. Les cèl·lules de la frontera del cervell
posterior s’especifiquen a la interfície entre compartiments adjacents. Les
cèl·lules de la frontera constitueixen funcions especialitzades i un perfil genètic
específic. Quan els compartiments adjacents es troben implicats en la
neurogènesi, les cèl·lules de la frontera es mantenen no neurogèniques. Per tal
d’esclarir com les cèl·lules de la frontera mantenen la seva pluripotència, hem
combinat estratègies funcionals amb la presa d’imatges d’alta resolució en viu
en embrions de peix zebra. A estadis temprans, les cèl·lules de la frontera
mostren un enriquiment de her9 i una absència en l’activitat de Notch.
Posteriorment, les cèl·lules de la frontera perden l’enriquiment i activen Notch3,
fet que dispara la transició de progenitors neuroepitelials a progenitors de la glia
radial que pateixen divisions asimètriques. Her9 manté les cèl·lules de la frontera
com a progenitors promovent divisions simètriques proliferatives i prevenint les
divisions asimètriques. A més, Her9 també controla potencialment la proliferació
de les cèl·lules de la frontera a través de la inhibició de cdkn1ca i del
manteniment de l’activitat de YAP/TAZ-TEAD. En conjunt, Her9 manté les
cèl·lules de la frontera del cervell posterior com progenitores durant els estadis
embrionaris temprans.Programa de doctorat en Biomedicin
Knowledge graph inference from text
No full textTreball fi de màster de: Master in Intelligent Interactive SystemsSupervisor: Maria Inés TorresKnowledge graphs (KGs) play a central role in representing structured information of real world entities and their relationships and supporting tasks such as reasoning, search, and question answering. Recent advances in natural language processing have opened the door to building KGs directly from text, making it possible to automate the extraction of entities and their semantic relations at scale. This work
presents a complete pipeline for constructing a knowledge graph from natural language and using it to improve the accuracy and reliability of answers generated by large language models (LLMs).
The proposed system consists of four main stages. First, entities are identified using a domain adapted Named Entity Recognition model trained with BIO tagging on a corpus of automatically generated sentences. Then, a relation extraction component generates subject–predicate–object triples using a generative Transformer based model fine-tuned with domain-specific data. These structured triplets are
encoded into dense vectors using semantic embeddings and indexed using a high performance HNSW (Hierarchical Navigable Small World) search structure. This allows efficient retrieval of relevant facts at inference time. To evaluate the performance of the system, experiments were conducted across twelve semantic domains using the WikiDialogue dataset. The evaluation focused on entity detection accuracy and triplet extraction precision. Results show that training on multiple domains significantly improves generalization. Moreover, the use of strict entity boundaries and span-based matching helps reduce false positives. In the final stage, the indexed triplets are used as input to an LLM (LLaMA 3.3–70B Instruct) using a Retrieval-Augmented Generation setup. Given a user query, the system retrieves semantically similar triplets and includes them in the model prompt. Three prompting strategies are explored: no context (the model answers based solely on its internal knowledge), strict (answers only if the fact is explicitly stated
in the retrieved context), and permissive (uses background knowledge when the context is insufficient). Examples show that strict prompts reduce hallucinations, while permissive prompts provide fluent answers but may introduce unsupported information. Overall, the results confirm that a modular KG-based pipeline can enhance LLM output by grounding it in structured knowledge. The system is scalable, adaptable to multiple domains, and highlights the trade-offs between precision, recall, and fluency in generation tasks. This opens the door to applying similar approaches in settings where precision and clarity are essential, such as healthcare, finance, or enterprise data systems
Evaluating the quality of real-world data on adherence to oral endocrine therapy in breast cancer patients: How real is real-world data?
No full textPurpose: The aim of this study was to compare estimates of adherence to oral endocrine therapy (OET) based on real-world data (RWD) and on clinical evaluation in people diagnosed with breast cancer in the public healthcare system in Catalonia (Spain). Methods: We conducted two retrospective cohort studies. Cohort 1 (RWD) consisted of women diagnosed with breast cancer in 2021 in the public healthcare system of Catalonia (Spain). Sources of RWD were the pharmacy billing register, hospital discharge records, and the Catalan health division's central insurance registry. Nonadherence was defined as below 80% adherence in the first year of treatment. Data for cohort 2 came from two population-based cancer registries in Girona and Tarragona (Catalonia), with diagnoses from 2007 to 2011. We evaluated the impact of variables missing from RWD, such as stage and hormonal status. Analyses were performed using a chi-square test and logistic regression, with results stratified by age group and drug type. Results: Nonadherence at one year was 10.9% in cohort 1 and 11.3% in cohort 2. When we reviewed the medical records of a selection of nonadherent women from cohort 1, we found only 59.4% had documented treatment interruptions. Reasons for interruptions in the patients from RWD cohort included adverse effects (48.8%), patient decision (40.0%), medical reasons (29.4%), and other clinical causes (14.7%). Women aged under 50 years and those receiving tamoxifen or a sequential regimen had lower adherence. Determinants associated with nonadherence were similar in both approaches used. Conclusions: This study confirms the validity of estimating adherence with RWD from the Spanish national health system, although when combined with reviewing medical records, this may provide more reliable and higher-quality data. The RWD method provides valuable evidence to help oncologists discuss adherence with their patients