Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 15
Encontrada(s) 1 página(s)

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Enhancing Precision Medicine: An Automatic Pipeline Approach for Exploring Genetic Variant-Disease Literature

  • Contreras-Ochando, Lidia
  • Pla Santamaría, Ferran
  • Segarra Soriano, Encarnación
  • Marco-García, Pere|||0009-0003-7026-3543
  • León-Palacio, Ana|||0000-0003-3516-8893
  • Hurtado Oliver, Lluis Felip|||0000-0002-1877-0455
[EN] Advancements in genomics have generated vast amounts of data, requiring efficient methods for exploring the relationships between genetic variants and diseases. This paper presents a pipeline approach that automatically integrates diverse biomedical databases, including NCBI Gene, MeSH, LitVar2, PubTator, and SynVar, for retrieving comprehensive information about genes, variants, diseases, and associated literature. The pipeline consists of multiple stages: querying and searching across the different databases, extracting relevant data, and applying filters to refine the results. Its goal is to bridge the gap in information retrieval related to genetic variants and diseases by providing a systematic framework for discovering relevant literature. The pipeline uses open-access sources to uncover additional articles not referenced in expert reports that mention the genetic variants of interest. In this paper, we present the methodology of the pipeline, discuss its limitations and highlight its potential for advancing information systems, data management, and interoperability in the domains of genomics and precision medicine., This work is partially supported by MCIN/AEI/10.13039/501100011033, by the 'European Union' and 'NextGenerationEU/MRR', and by 'ERDF A way of making Europe' under grants PDC2021-120846-C44 and PID2021-126061OB-C41. It is also partially supported by the Generalitat Valenciana underproject CIPROM/2021/023. We would like to thank the authors of LitVar2 for their valuable assistance.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

CLARA: Semi-automatic Retraining System

  • Campos-Mocholí, Mario|||0009-0009-4721-2207
  • Chacón-Albero, Oriol de la Pau
  • Julian, Vicente|||0000-0002-2743-6037
  • Botti V.|||0000-0002-6507-2756
  • Marco-Detchart, Cédric
  • Rincón-Arango, Jaime Andrés
[EN] Recent advancements in computer vision have led to humanlevel performance in image recognition tasks, but challenges persist in real-world applications due to differences in data distribution. This paper introduces CLARA, a semi-automatic framework designed to continually retrain computer vision models by incorporating expert annotation of new data. The system was tested in a production environment with a waste image classification model and showed improved performance and robustness against evolving data distributions. This approach encourages collaboration among stakeholders in waste management and utilizes human-in-the-loop strategies to enhance model accuracy. The framework is adaptable to various models and tasks and, additionally, supports fusion techniques that combine outputs from multiple models to improve prediction robustness and overall system reliability., This work was partially supported with grant tPID2021- 123673OB-C31 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe , grant CIPROM/2021/077 from the Conselleria de Innovació, Universitats, Ciencias i Societat Digital GVA - PROMETEO, grant PAID-06-23 by the Vice Rectorate Office for Research from UPV and project Reciclaje Inteligente y Cooperativo en Toda la Cadena de Valor Orientado a Una Sociedad Sostenible 360º (INNEST/2022/180) from AVI.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Using Diffusion Models for Data Augmentation on Limited Rodent OCT Datasets

  • García-Torres, Fernando|||0000-0002-5337-8774
  • del Amor, Rocío
  • Morales, Sandra|||0000-0003-0763-1545
  • Naranjo Ornedo, Valeriana|||0000-0002-0181-3412
  • Barroso, Álvaro
  • Kemper, Björn
  • Schnekenburger, Jürgen
[EN] Optical coherence tomography (OCT) has transformed ophthalmology by enabling accurate diagnosis of retinal diseases. OCT in rodents is used in research studies of ophthalmic diseases and treatment evaluation. However, the scarcity of data limits the effectiveness of deep learning models for this purpose. This study explores the use of synthetic data generation to improve deep learning models for OCT image analysis from rodent retinas. To address this, the research leverages advanced class-conditional generative models to create high-quality synthetic images of ex-vivo rodent retinas embedded in two different mediums (solid resin and liquid media) with a reduced initial database. The conditional Denoising Diffusion Probabilistic Model (DDPM) proved to be the best model, as it produces more realistic images that closely mimic the characteristics of real rodent retinal scans. By augmenting the training datasets with these synthetic images, the performance of deep learning models in segmenting retinal layers improved by 7%. This approach reduces reliance on scarce biological samples and animal testing, promoting ethical research practices while improving the accuracy and efficiency of AI-driven medical diagnostics. The integration of synthetic data generation represents a significant advance in medical imaging, offering a scalable solution to data scarcity, generating reliable samples in two mediums and promoting the development of more robust and accurate diagnostic tools., This work has received funding from Horizon Europe, the European Union s Framework Programme for Research and Innovation, under Grant Agreement No. 101070062 (SEQUOIA). Generalitat Valenciana partially funded this work through project CIPROM/2022/20. The work of Sandra Morales has been cofunded by the Universitat Polit`ecnica de Val`encia, Spain through the program PAID10-20. The work of Rocio del Amor has been supported by the Spanish Ministry of Universities under an FPU Grant (FPU20/05263).
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

An Adaptive LoRa-Based Edge Architecture for Data Sensing and Collection in Natural Parks

  • Cecilia-Canales, José María|||0000-0001-5648-214X
  • Arratia-Uribe, Benjamín Andrés|||0000-0001-7869-053X
  • Rosas-Olivos, Erika Susana
  • Martín Furones, Ángel Esteban|||0000-0001-9379-0694
  • Perles, Angel|||0000-0002-5782-4284
  • Posadas-Yagüe, Juan-Luis|||0000-0002-5017-8689
  • Manzoni, Pietro|||0000-0003-3753-0403
[EN] This article introduces the design of a flexible IoT/Edge architecture to overcome the challenges of deploying advanced sensors in remote settings such as natural parks. These environments demand reliable communication and considerable computational resources to manage the extensive real-time data from IoT devices like sensors and cameras. Our approach combines two LoRa-based protocols to efficiently handle large data blocks and ensure long-range communication, conserving energy and maintaining data integrity crucial for integrating sophisticated sensors with conventional IoT devices. The architecture offers edge computing within a containerbased gateway, simplifying on-site data processing and minimizing data traffic complexities. The MQTT protocol supports interoperability among distributed services, enhancing adaptability and efficiency in processing data locally or via cloud services like digital twins. We demonstrate this system¿s effectiveness through two case studies in natural parks: visitor counting using Time of Flight technology and water level measurement with GNSS data. These cases highlight the system¿s ability to handle frequent small and less frequent large data transmissions efficiently, This work has been supported by the MCSA-SE REMARKABLE grant ID: 101086387; Research project CIPROM/2023/29, funded by Direccio General de Ciencia ` i Investigacio Generalitat Valenciana - SPAIN; EU s Horizon ¿ 2020 research and innovation program No 101017861, by project Ramon y Cajal Grant RYC2018-025580-I, funded by ¿ MCIN/AEI/ 10.13039/501100011033 and ESF Investing in your future, GVA GRISOLIAP/2021/103 and Grant TED2021- 130890B funded by MCIN/AEI and by EU NextGenerationEU/PRTR.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Towards a Low-Rank Approach to Compress Deep Neural Networks

  • Liern-Garcia, Marina
  • López-García, Aaron
  • Marco-Detchart, Cédric
  • Carrascosa Casamayor, Carlos|||0000-0003-3649-6530
[EN] In this work, we introduce a low-rank approach based on the truncated Singular Value Decomposition (SVD) technique to make deep neural networks (DNNs) smaller. Our method focuses on reducing the overall physical size of a neural network model without losing much in terms of its accuracy or increasing its error noticeably. Specifically, we applied our technique to a modified MobileNetV2 model, in the context of automatic plant diseases detection. We chose two of the biggest layers of the model and compressed them. Our goal was to see how this compression affects the breakdown of the layers, the error in rebuilding the layers, and how well the modified model performs. The results showed that the smaller model still predicts very accurately, even with the reduced physical size of its layers. By making these layers smaller, our approach offers a practical way to handle the deployment of large neural networks, especially in devices with limited resources, without compromising their effectiveness., This work was partially supported with grants TED2021-131295B-C32, PID2021-123673OB-C31 funded by MCIN/AEI/ 10.13039/ 501100011033 and by ERDF A way of making Europe , PROMETEO grant CIPROM/2021/077 from the Conselleria de Innovación, Universidades, Ciencia y Sociedad Digital - Generalitat Valenciana, Early Research Project grant PAID-06-23 by the Vice Rectorate Office for Research from Universitat Politècnica de València (UPV)
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Towards Sustainable Recycling: Advancements in AI-Based Waste Classification

  • Chacón-Albero, Oriol de la Pau
  • Campos-Mocholí, Mario|||0009-0009-4721-2207
  • Julian, Vicente|||0000-0002-2743-6037
  • Botti V.|||0000-0002-6507-2756
  • Marco-Detchart, Cédric
  • Rincón-Arango, Jaime Andrés
[EN] Despite global efforts in the pursuit of sustainable waste management practices, significant gaps for improvement persist, as evidenced by Spain¿s recycling rates lagging below European targets. As part of a larger initiative aiming to build an intelligent and collaborative environment to promote the Circular Economy, this paper investigates the potential of Image Classification in developing an intelligent assistant for waste disposal guidance. Tailored to Spain¿s recycling framework, an analysis of different state-of-the-art Image Classification architecture¿s performance on waste sorting is presented. Additionally, the paper explores aggregation techniques to enhance classification accuracy., This work was partially supported with grant PID2021- 123673OB-C31, TED2021-131295B-C32 funded by MCIN/AEI/10.13039/5011 00011033 and by ERDF A way of making Europe , grant CIPROM/2021/077 from the Conselleria de Innovació, Universitats, Ciencias i Societat Digital Generalitat Valenciana - PROMETEO and project Reciclaje Inteligente y Cooperativo en Toda la Cadena de Valor Orientado a Una Sociedad Sostenible 360º (INNEST/2022/180) from Agencia Valenciana de la Innovación.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

ELiRF-VRAIN at BioNLP Task 1B: Radiology Report Summarization

  • Ahuir-Esteve, Vicent|||0000-0001-5636-651X
  • Hurtado Oliver, Lluis Felip|||0000-0002-1877-0455
  • Segarra Soriano, Encarnación
[EN] This paper presents our system at the Radiology Report Summarization Shared Task-1B of the 22nd BioNLP Workshop 2023. Inspired by the work of the BioBART model, we continuously pre-trained a general domain BART model with biomedical data to adapt it to this specific domain. In the pre-training phase, several pre-training tasks are aggregated to inject linguistic knowledge and increase the abstractivity of the generated summaries. We present the results of our models, and also, we have carried out an additional study on the lengths of the generated summaries, which has provided us with interesting information., This work is partially supported by MCIN/AEI/10.13039/501100011033, by the "European Union and 'NextGenerationEU/MRR', and by 'ERDF A way of making Europe' under grants PDC2021-120846-C44 and PID2021-126061OB-C41. It is also partially supported by the Generalitat Valenciana under project CIPROM/2021/023, and by the Spanish Ministerio de Universidades under the grant FPU21/05288 for university teacher training.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Protocol Dialects as Formal Patterns

  • Galán, D.
  • Meadows, C.
  • Meseguer, J.
  • García-Valero, Víctor|||0000-0003-0681-1130
  • Escobar Román, Santiago|||0000-0002-3550-4781
[EN] A protocol dialect is a lightweight method to obfuscate the communication exchanges between legitimate protocol users to make it hard for malicious users to interact with legitimate ones. So far, dialects have been based on a single obfuscation transformation, which we call a lingo. In this work dialects are generalized to become protocol and lingo generic. In this way they can be composed with other dialects for greater security. We present a formal semantics of dialects as formal patterns, i.e., as protocol transformations formalized in rewriting logic. We also present several attacker models and explain how our generalized notion of dialect can be used to harden dialects against such attackers., S. Escobar and V. Garcia have been partially supported by the grant PID2021-122830OB-C42 funded by MCIN/AEI/10.13039/501100011033 and ERDF A way of making Europe, by the grant CIPROM/2022/6 funded by Generalitat Valenciana, and by the grant PCI2020-120708-2 funded by MICIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR. D. Galan, V. Garcia, C. Meadows, and J. Meseguer have been partially supported by ONR Code 311.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Estudio con simulaciones Monte Carlo del espectro de neutrones en una sala de tratamiento de protonterapia

  • Díaz-Comeche, A.
  • Oliver-Gil, Sandra|||0000-0001-8258-3972
  • Juste-Vidal, Belen-Jeanne|||0000-0003-1978-3765
  • Miró Herrero, Rafael|||0000-0003-1012-0869
  • Verdú Martín, Gumersindo Jesús|||0000-0001-5098-080X
[ES] La protonterapia es una técnica radioterapéutica que consiste en acelerar protones hasta energías de entre 70 y 250 MeV. Su principal ventaja frente a las terapias que emplean fotones o electrones es que la dosis radiactiva se deposita de manera más localizada, por lo que aparecen menos efectos secundarios. Durante este proceso se generan neutrones con energías de hasta 230 MeV, como producto de reacciones (p,n) en el cabezal del acelerador. Estos neutrones contribuyen a la dosis recibida por el paciente, pero normalmente no se tienen en cuenta al planificar el tratamiento. Determinar el espectro energético de los neutrones es fundamental debido a que la dosis que depositan depende, en mayor medida, de su energía. En este trabajo se ha empleado un algoritmo MLEM (Maximum Likelihood Expectation-Maximization) para determinar el espectro neutrónico producido en un cabezal de protonterapia. Dicho espectro se ha reconstruido a partir de simulaciones Monte Carlo en MCNP6.3 que recrean las medidas obtenidas por un espectrómetro de esferas Bonner situado en la salida del cabezal. Las simulaciones Monte Carlo realizadas sirven como primer paso para la medición de un espectro de neutrones experimental en una instalación de protonterapia, así como para la futura validación de modelos que intenten representar las condiciones propias de dichas instalaciones. El espectro obtenido mediante el procedimiento descrito se ajusta al proporcionado por otros trabajos publicados previamente, en los cuales también se han empleado métodos de Monte Carlo. Así, queda demostrada la fiabilidad de la metodología seguida. Se ha estudiado también la variación del espectro resultado de cambiar los parámetros de entrada del algoritmo MLEM, determinando los más eficaces para reconstrucción de espectros de estas características. Así, con este trabajo se abre la puerta a futuras investigaciones en el área de la protonterapia, técnica de especial interés y que ha experimentado un gran auge en los últimos años.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Calidad del agua en embalses de aguas claras de la cuenca del Júcar con imágenes Sentinel-2 nivel L2A

  • Caballero, Gabriel
  • Delegido, Jesus
  • Sòria-Perpinyà, Xavier
  • Sánchez-Pérez, Lissette
  • Ruiz-Verdú, Antonio
  • Alvado, Bárbara
  • Durán, María Concepción
  • Ortega, Danián
  • Vicente, Eduardo
  • Moreno, José
[ES] Sentinel-2 es una misión satelital del programa Copernicus que ofrece imágenes de 10 m de resolución espacial cada cinco días. Numerosos estudios han demostrado que es adecuado para la estimación de calidad de aguas en lagos y embalses. En este trabajo se estudian 7 embalses españoles de aguas claras de la cuenca del Júcar (Alarcón, Benagéber, Contreras, María Cristina, Regajo, Sitjar y Tous) en los que se han medido distintas variables, entre otras, la clorofila-a (Chl-a), los sólidos en suspensión (TSS), la profundidad de Disco de Secchi (SDD) y la ficocianina (PC). Con estos datos, junto a los espectros de reflectividad obtenidos de imágenes Sentinel-2 nivel L2A, es decir, con la corrección atmosférica que realiza la Agencia Espacial Europea, se han estudiado diferentes índices espectrales. Los resultados muestran que la PC se puede estimar con un índice NDI con las bandas B3 (560 ±17 nm) y B2 (490 ±32 nm), y un ajuste lineal; el SDD se puede obtener mediante un ajuste exponencial de un índice que combina las bandas B2, B3 y B4 (665 ±15 nm) y los TSS se pueden estimar con una relación lineal con un índice que combina las bandas B1 (443 ±10 nm), B2 y B3. Sin embargo, para la Chl-a no se ha encontrado un modelo único para todos los embalses, por lo que se han separado en dos grupos en función de su respuesta espectral. Para un grupo de embalses el modelo que da mejores resultados usa una combinación lineal de B2, B3 y B4 y en el otro se usa un índice tribanda, con las mismas bandas. Estos modelos permiten estimar la calidad del agua de estos embalses, mostrando tanto su distribución espacial como series temporales. Todo ello con un procesado sencillo y rápido que se puede hacer en la nube con las herramientas gratuitas de Copernicus., Este trabajo se ha llevado al cabo en el marco de dos proyectos de investigación. Por una parte, el convenio entre la Confederación Hidrográfica del Júcar y la Universidad de Valencia para el seguimiento de la calidad de aguas en embalses, lagos y otras masas de agua de la cuenca del Júcar, a partir de datos de sensores remotos de observación de la Tierra (Convenio TELEDEMB23). Por otra parte, el proyecto RESSBIO (Remote Sensing Spectroscopy for wetlands Biodiversity, CIPROM2021049) subvencionado por la Generalitat Valenciana dentro del programa Prometeo.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Personal Data Flows and Privacy Policy Traceability in Third-party LLM Apps in the GPT Ecosystem

  • Carrillo-Moraga, Juan Carlos
  • Martin-Navarro, Jose Luis|||0000-0002-4503-4189
  • Such Aparicio, José Miguel|||0000-0002-6041-178X
  • Ma, Rongjun
[EN] The rapid growth of platforms for customizing Large Language Models (LLMs), such as OpenAI's GPTs, has raised new privacy and security concerns, particularly related to the exposure of user data via third-party API integrations in LLM apps. To assess privacy risks and data practices, we conducted a large-scale analysis of OpenAI¿s GPTs ecosystem. Through the analysis of 5,286 GPTs and the 44,102 parameters they use through API calls to external services, we systematically investigated the types of user data collected, as well as the completeness and discrepancies between actual data flows and GPTs' stated privacy policies. Our results highlight that approximately 35% of API parameters enable the sharing of sensitive or personally identifiable information, yet only 15% of corresponding privacy policies provide complete disclosure. By quantifying these discrepancies, our study exposes critical privacy risks and underscores the need for stronger oversight and support tools in LLM-based application development. Furthermore, we uncover widespread problematic practices among GPT creators, such as missing or inaccurate privacy policies and a misunderstanding of their privacy responsibilities. Building on these insights, we propose design recommendations that include actionable measurements to improve transparency and informed consent, enhance creator responsibility, and strengthen regulation., This research was partially funded by the INCIBE s strategic SPRINT (Seguridad y Privacidad en Sistemas con Inteligencia Artificial) C063/23 project with funds from the EU-NextGenerationEU through the Spanish government s Plan de Recuperación, Transformación y Resiliencia; by the Spanish Government via project PID2023- 151536OB-I0; and by the Generalitat Valenciana via project PROMETEO CIPROM/2023/23.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Distilling the Effects of Language Model Contamination

  • Mehrbakhsh, Behzad|||0000-0001-9017-989X
  • Martínez-Plumed, Fernando|||0000-0003-2902-6477
  • Hernández-Orallo, José|||0000-0001-9746-7632
[EN] The proportion of AI-generated content permeating the well of knowledge is increasing significantly. Large language models (LLMs) contribute to that contamination but they also suffer from it. However, it is yet to be clarified the effect of different sources of error, be it human-generated or LLM-generated. Controlling for the percentage of error, we explore the impact on LLM fine-tuning when errors come from humans, from other language models or are generated randomly using an aleatoric or epistemic source. In this paper, we compare these different types of error for in-distribution and out-of-distribution experimental settings. By analysing the levels of errors and their distribution, we find a nuanced view: while in-distribution human-generated noise seems more benign than the LLM-generated counterpart, in the out-of-distribution case the model-generated noise may not be necessarily worse., We thank the anonymous reviewers for their comments and Dr. Peter Clark for providing us with the extended version of the OpenBookQA dataset. This work was funded by valgrAI, the Future of Life Institute, FLI, under grant RFP2-152, the EU (FEDER) and Spanish grant RTI2018-094403-B-C32 funded by MCIN/AEI/10.13039/501100011033 and by CIPROM/2022/6 funded by Generalitat Valenciana, EU s Horizon 2020 research and innovation programme under grant agreement No. 952215 (TAILOR), and Spanish grant PID2021-122830OB-C42 (SFERA) funded by MCIN/AEI/10.13039/501100011033 and "ERDF A way of making Europe"
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

The use of podcasting for a hybrid flipped classroom

  • Marco-Detchart, Cédric
  • Taverner-Aparicio, Joaquín José|||0000-0002-5163-5335
  • Jordán, Jaume|||0000-0003-0400-9136
[EN] We are currently assisting a social paradigm change motivated by the incorporation, more and more accelerated, of new information technologies (e.g., social networks or content platforms) in our day-today life. The increased availability of online resources in a universal, diverse, and permanent way is modifying how we consume online information and content. This is especially true when it comes to children and teenagers. New generations are increasingly adapting to immediacy and communication through information technologies. Therefore, it is necessary to evolve the educational paradigm to adapt it to this new social reality. Future learning strategies should consider the latest models of social communication, adapting them to achieve learning objectives from the perspective of constructive alignment and the acquisition or improvement of transversal competencies. In this sense, there are currently several technologies that can be incorporated into the classroom. One of the emerging technologies is the podcast. Usually used for entertainment (e.g., stories, books, or radio talks), the podcast is becoming a tool for massive online content distribution. There exist many advantages to using podcasting for an educative purpose, as the production cost (in terms of time) is lower than recording a video. From a content consumer perspective, the main advantage is that an audio-only approach can be used/consumed everywhere and more easily than video media. Some educational podcasts are available online but generally tend to focus on learning languages or history. Outside these specific topics, their use in technology subjects is still residual and mainly focuses on interviews or long expositions. Moreover, in current proposals, the teacher is the one who produces the podcast, and therefore it is a one-way communication model (from the teacher to the students). On the other hand, one of the teaching innovation models that is being increasingly used is the flipped classroom. In the flipped classroom, students are the main protagonists in their learning. They must prepare the theoretical parts on their own, and the teacher serves as a guide during this learning process. In this paper, we propose using the podcast in the flipped classroom model, turning the podcast into a bidirectional tool in which students are both producers and receivers of learning. The proposal consists of dividing the class into small groups of students (2-4). Each group will record a podcast episode on a different conceptual part of the lesson. Group members will have to coordinate and share the activities of recording and searching for content for the podcast. This will encourage transversal skills related to communication and social skills. These podcasts will then be shared with the rest of the groups so that everyone can have direct and permanent access to the different sections of the lesson. Creativity will also be encouraged, allowing students to add the music or sound effects they consider necessary to enhance the explanation., The authors gratefully acknowledge the financial support of Consellería d'Innovació, Universitats, Ciencia i Societat Digital from Comunitat Valenciana and the European Social Fund (Investing In Your Future) (APOSTD/2021/227 and CIPROM/2021/077), the Spanish Ministry of Science (project PID2021-123673OB-C31) and the Research Services of Universitat Politècnica de València. Jaume Jordán is supported by grant IJC2020-045683-I funded by MCIN/AEI/10.13039/501100011033 and by "European Union NextGenerationEU/PRTR".
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Evaluating the use of self-video teaching in a flipped classroom

  • Taverner-Aparicio, Joaquín José|||0000-0002-5163-5335
  • Jordán, Jaume|||0000-0003-0400-9136
  • Vos, Tanja Ernestina|||0000-0002-6003-9113
  • Marco-Detchart, Cédric
[EN] New generations are increasingly becoming more familiar with consuming audio-visual material through online platforms. Consequently, learning through IT-based tools is becoming more and more common. Nowadays, learning platforms (e.g., edX or W3Schools) or content platforms (e.g., YouTube) containing vast amounts of courses and video tutorials are becoming increasingly popular among students. The main advantage of online learning is that students can access the content from anywhere and whenever they want, being able to revisit the content to review concepts and improve their level of knowledge. In this way, learning based on a deep approach and self-learning is promoted since students are the ones who regulate their learning process by deciding how much time to dedicate and when to do it. Appropriately using this type of resource can become a very effective tool applied to a flipped classroom model. In the flipped classroom model, students are active learners since they are in charge of developing the lesson material both in class and at home. In this type of learning, the teacher assumes the role of guide assisting during the learning process. A standard methodology in this flipped classroom model consists of students preparing different parts of the course content and then explaining those parts to their classmates. In this way, students develop a sense of responsibility toward the rest of their classmates, creating an environment where they can recognise their shortcomings and take control of their learning to teach others. In addition, the acquisition of transversal communication skills is encouraged. With all this in mind, in this article, we describe a case study we are currently carrying out with students enrolled in the programming course at the Universitat Politècnica de València. Our proposal combines the flipped classroom model with access to online resources. In this first approach, we have proposed that the students record a video explaining a part of the lesson or how to solve at least two exercises step by step. The explanation must be done as if they were content creators, and their audience were beginner programmers. The students will upload the videos to a private YouTube channel that will only be accessible to their classmates. In the classroom, the teacher will encourage students to share their stories and experiences while learning, editing, and recording the videos. This proposal's main objective is to promote students' engagement in the learning process and offer them learning alternatives through online content with a closer language that they can access whenever they need it. To motivate participation, students and teachers will choose the three best videos from all the videos. The three winners will receive extra points in the evaluation of the course., The authors gratefully acknowledge the financial support of Consellería d'Innovació, Universitats, Ciencia i Societat Digital from Comunitat Valenciana and the European Social Fund (Investing In Your Future) (APOSTD/2021/227 and CIPROM/2021/077), the Spanish Ministry of Science (project PID2021-123673OB-C31) and the Research Services of Universitat Politècnica de València. Jaume Jordán is supported by grant IJC2020-045683-I funded by MCIN/AEI/10.13039/501100011033 and by "European Union NextGenerationEU/PRTR".
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Language Task Difficulty Prediction through LLM-Annotated Meta-Features

  • Moros-Daval, Yael
  • Martínez-Plumed, Fernando|||0000-0003-2902-6477
  • Hernández-Orallo, José|||0000-0001-9746-7632
[EN] Assessing the capabilities of large language models (LLMs) is increasingly challenging due to their generality and uneven task performance. Often, we do not know how much of the success or failure on a particular task is due to the `loading¿ of the language elements in the task, such as narrative understanding, or some other intrinsic (non-linguistic) components, such as domain-specific common sense or reasoning capabilities. Understanding what tasks are most loaded on language and determine the predictability of LLMs on these tasks is crucial for improving benchmarks, designing better LLMs, and ensuring their safe deployment. We present an innovative methodology that uses LLMs to annotate linguistic metafeatures , allowing us to predict task difficulty and understand linguistic loadings more accurately than traditional readability scores. Using GPT-4 for automated annotation, we show strong predictability for a variety of tasks and language models (e.g., MMLU with R2 from 0.68 to 0.83), but observe limited predictability for other tasks (e.g., LSAT with R2 of -0.07)., We acknowledge support from: Grant for Master Studies funded by ValgrAI Valencian Graduate School and Research Network for Artificial Intelligence and Generalitat Valenciana; FISCALTICS (I+D+i PID2022-140110OA-I00), granted by MICIU/AEI/10.13039/ 501100011033 and by ERDF, EU; CIPROM/2022/6 (FASSLOW) and IDIFEDER/2021/05 (CLUSTERIA) funded by Generalitat Valenciana; the EC H2020-EU grant agreement No. 952215 (TAILOR); US DARPA HR00112120007 (RECoG-AI) and Spanish grant PID2021-122830OB-C42 (SFERA) funded by MCIN/AEI/10.13039/501100011033 and ERDF.