Untitledhttps://hdl.handle.net/10259/73492026-05-11T23:39:56Z2026-05-11T23:39:56ZDetection of Stress Stimuli in Learning Contexts of iVR EnvironmentsRamírez Sanz, José MiguelPeña-Alonso, Helia MarinaSerrano Mamolar, AnaArnaiz González, ÁlvarBustillo Iglesias, Andréshttps://hdl.handle.net/10259/109062025-10-01T00:05:31Z2023-09-05T00:00:00ZDetection of Stress Stimuli in Learning Contexts of iVR Environments
Ramírez Sanz, José Miguel; Peña-Alonso, Helia Marina; Serrano Mamolar, Ana; Arnaiz González, Álvar; Bustillo Iglesias, Andrés
The use of eye-tracking in immersive Virtual Reality (iVR) is becoming an important tool for improving the learning outcomes. Nevertheless, the best Machine Learning (ML) technologies for the exploitation of eye-tracking data is yet unclear. Actually, one of the main drawbacks of some ML technologies, such as classifiers, is the scarce labeled data for training models, being the process of data annotation time-consuming and expensive. This paper presents a complete experimentation where different ML algorithms were tested, both supervised and semi-supervised, for trying to identify the stressors/distractors present in iVR learning experiences simulating the operation of a bridge crane. Results shown that the use of semi-supervised techniques can improve the performance of the Machine Learning methods making possible the identification of stressful situations in iVR environments. The use of semi-supervised learning techniques makes possible training ML algorithms without the need of great amount of labeled data which makes the data exploitation cheaper and easier.
Comunicación presentada en: International Conference on Extended Reality, XR Salento 2023, held in Lecce, Italy during September 6–9, 2023
2023-09-05T00:00:00ZEvaluation of Semi-Supervised Machine Learning applied to Affective State DetectionMartin-Melero, ÍñigoSerrano Mamolar, AnaRodríguez Diez, Juan Joséhttps://hdl.handle.net/10259/109022026-04-22T22:42:17Z2024-04-23T00:00:00ZEvaluation of Semi-Supervised Machine Learning applied to Affective State Detection
Martin-Melero, Íñigo; Serrano Mamolar, Ana; Rodríguez Diez, Juan José
The affective computing field usually concerns data that is difficult, expensive or time-consuming to label. One way to overcome this limitation is the application of Semi-Supervised Machine Learning, that typically works with a small set of labeled data and a larger one of unlabeled data. This paper assesses the suitability of these techniques on the prediction of affective state, by analyzing the physiological and emotional response data of 30 different subjects while watching several emotion-eliciting videos. Three Semi-Supervised Learning algorithms are compared with their Supervised base classifiers in both a subject-independent and subject-dependent analyses, across a widely extended dataset. In view of the results, it can be concluded that Semi-Supervised Learning did not outperform their respective Supervised base classifiers for this specific dataset as it was expected. Subject-dependent classification resulted in accuracy rates between 68% and 85%, whereas the accuracy rates were between 38% and 72% for subject-independent classification.
Comunicación presentada en: EmotionAware 2024: Eighth International Workshop on Emotion Awareness for Pervasive Computing Beyond Traditional Approaches, held as part of the 2024 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops), 11–15 March 2024, Biarritz, France.
2024-04-23T00:00:00ZPersonalising the Training Process with Adaptive Virtual Reality: A Proposed Framework, Challenges, and OpportunitiesLucas Pérez, GadeaRamírez Sanz, José MiguelSerrano Mamolar, AnaArnaiz González, ÁlvarBustillo Iglesias, Andréshttps://hdl.handle.net/10259/109012025-09-30T00:05:36Z2024-09-11T00:00:00ZPersonalising the Training Process with Adaptive Virtual Reality: A Proposed Framework, Challenges, and Opportunities
Lucas Pérez, Gadea; Ramírez Sanz, José Miguel; Serrano Mamolar, Ana; Arnaiz González, Álvar; Bustillo Iglesias, Andrés
This work presents a conceptual framework that integrates Artificial Intelligence (AI) into immersive Virtual Reality (iVR) training systems, aiming to enhance adaptive learning environments that dynamically respond to individual users’ physiological states. The framework uses real-time data acquisition from multiple sources, including physiological sensors, eye-tracking and user interactions, processed through AI algorithms to personalise the training experience. By adjusting the complexity and nature of training tasks in real time, the framework seeks to maintain an optimal balance between challenge and skill, fostering an immersive learning environment. This work details some methodologies for data acquisition, the preprocessing required to synchronise and standardise diverse data streams, and the AI training techniques essential for effective real-time adaptation. It also discusses logistical considerations of computational load management in adaptive systems. Future work could explore the scalability of these systems and their potential for self-adaptation, where models are continuously refined and updated in real-time based on incoming data during user interactions.
Comunicación presentada en: International Conference on Extended Reality, XR Salento 2024, held in Lecce, Italy during September 4–7, 2024
2024-09-11T00:00:00ZTowards Automatic Tutoring of Custom Student-Stated Math Word ProblemsArnau-González, PabloSerrano Mamolar, AnaKatsigiannis, StamosArevalillo-Herráez, Miguelhttps://hdl.handle.net/10259/109002025-09-30T00:05:32Z2023-06-30T00:00:00ZTowards Automatic Tutoring of Custom Student-Stated Math Word Problems
Arnau-González, Pablo; Serrano Mamolar, Ana; Katsigiannis, Stamos; Arevalillo-Herráez, Miguel
Math Word Problem (MWP) solving for teaching math with Intelligent Tutoring Systems (ITSs) faces a major limitation: ITSs only supervise pre-registered problems, requiring substantial manual effort to add new ones. ITSs cannot assist with student-generated problems. To address this, we propose an automated approach to translate MWPs to an ITS’s internal representation using pre-trained language models to convert MWP to Python code, which can then be imported easily. Experimental evaluation using various code models demonstrates our approach’s accuracy and potential for improvement.
Comunicación presentada en: 24th International Conference on Artificial Intelligence in Education, AIED 2023, Tokyo, Japan, July 3–7, 2023.
2023-06-30T00:00:00Z