Javier Sevilla Salcedo

Javier Sevilla Salcedo

I'm Javier Sevilla Salcedo, a Systems Engineering and Automation researcher at Carlos III University of Madrid currently pursuing a Ph.D. in Electrical Engineering, Electronics, and Automation. With a B.Sc. in Industrial Electronics and Automation Engineering from the University of Jaén, and a subsequent master's in Robotics and Automation at Carlos III.

My career spans practical experience in the automotive industry's robotics and automation sectors, along with over four years of research in process automation, social robotics, and Natural Language Processing (NLP). I designed a robotic guitar tuner for my bachelor's project and developed intuitive NLP applications for social robots during my master's. Currently, my research is deeply rooted in AI and NLP, with significant interests in Social Robotics and cutting-edge methodologies including Deep Learning and Deep Reinforcement Learning. My goal is to elevate the ways in which robots cognitively perceive and engage with the environment around them.

Journal Publications

Using Large Language Models to Shape Social Robots’ Speech

Using Large Language Models to Shape Social Robots’ Speech

Using Large Language Models to Shape Social Robots’ Speech [IJIMAI 8 (3)]

Two methods are proposed: one for rephrasing existing utterances to retain their meaning but introduce variation, and the other for creating topic-specific speech that adapts to the conversation partner. The results are promising, but challenges such as computational demands, interaction delays, reliance on proprietary models, and the need for subjective evaluation are noted as constraints.

Read Publication
Publication 8

A Bio-Inspired Endogenous Attention-Based Architecture for a Social Robot

A Bio-Inspired Endogenous Attention-Based Architecture for a Social Robot [Sensors 22 (14)]

Introducing a perception architecture for social robots, inspired by endogenous attention mechanisms, enhancing robots' interaction with their environment and users.

Read Publication

International Conferences

Publication 1

GERT: Transformers for Co-speech Gesture Prediction in Social Robots

GERT: Transformers for Co-speech Gesture Prediction in Social Robots [ICSR '23]

GERT (Gesture-Enhanced Robotics Transformer), a set of three transformer-based models designed to predict gestures that align with a robot's verbal communication. The research evaluates and compares the effectiveness and efficiency of the three model variations, demonstrating that they all perform well (with F-scores ranging from 0.78 to 0.86) while highlighting the balance between model size and performance.

Read Publication
Publication 4

User-Adapted Semantic Description Generation Using Natural Language Models

User-Adapted Semantic Description Generation Using Natural Language Models [ISAMI '22]

Exploring the creation of user-profile-specific descriptions by employing natural language generation models to provide more fluid and natural human-computer interactions.

Read Publication
Publication 5

Using Deep Learning for Implementing Paraphrasing in a Social Robot

Using Deep Learning for Implementing Paraphrasing in a Social Robot [ISAMI '22]

This research enhances the conversational abilities of a social robot by introducing models capable of generating contextually appropriate Spanish paraphrases.

Read Publication

National Conferences

Publication 3

Predicción de gestos no-verbales usando aprendizaje profundo

Predicción de gestos no-verbales usando aprendizaje profundo [XLIV Jornadas de Automática]

This study introduces a deep learning system using transformer models to predict non-verbal gestures for social robots, surpassing previous techniques for natural human-robot interactions.

Read Publication
Publication 6

ARTETERAPIA: una estrategia para incrementar el vínculo en la robótica social

ARTETERAPIA: una estrategia para incrementar el vínculo en la robótica social [XLIII Jornadas de Automática]

Addressing the societal divide between the elderly and the young, this article proposes using art therapy with social robots to enhance the psychophysical state of older adults.

Read Publication
Publication 7

Modelos de lenguaje natural para robots sociales

Modelos de lenguaje natural para robots sociales [XLIII Jornadas de Automática]

Discussing the role of natural language generation models in enhancing communication of social robots and making human-robot interactions more dynamic and adaptable.

Read Publication
Publication 9

Aplicando de estrategias psicológicas y sociales para incrementar el vínculo en interacción humano-robot

Aplicando de estrategias psicológicas y sociales para incrementar el vínculo en interacción humano-robot [Jornadas de Robótica, Educación y Bioingeniería 2022]

Enhancing the bond in human-robot interaction via social psychology and game theory principles, aiming to build trust and increase the usability of social robots.

Read Publication