This taskbot's vision is to move multimodal conversational AI closer to the way humans communicate and solve tasks.
Conversational assistants are constantly pushed to their limits supporting human daily routines. The constant demand to extend the reach of conversational assistants to many activities and tasks exposes the lack of General AI algorithms. Hence, our long-term vision is to move multimodal conversational AI closer to the way humans communicate and solve tasks. To achieve our long-term vision, we aim to make multimodal conversations human-like, stimulating, and resilient to user ramblings.
Rafael Ferreira - Team leader
Ferreira is a 2nd Year Computer Science PhD student at NOVA School of Science and Technology in Portugal. He concluded his master's thesis in 2021 in conversational search, and is currently working on the subject of dialogue policy for conversational task-oriented agents. With his research, he aims to improve the flow and naturalness of conversational agents with the aim of developing systems that exhibit a more human-like understanding. His interests include Conversational Agents, NLP, ML, and Multimodal-AI, with experience in working with Alexa and AWS from the previous TaskBot Edition.
Silva is a Ph.D. student on language generation for recommendation systems with a focus on politeness and empathy generation. Throughout his MSc and early stages of my Ph.D., he has acquired an extensive knowledge of state-of-the-art text generation and realization approaches. In 2019, he completed an internship where he integrated a team management platform with Alexa allowing team members to use their Echo devices to interact with the platform. In 2022 he was part of team TWIZ which won 2nd place in the Taskbot Challenge. His main interest remains in bridging the gap between human-machine communication towards creating natural conversational agents.
Tavares is currently beginning the second year of his PhD, focusing on Natural Language Understanding in conversational agents. Currently, his research focus has been in zero and few-shot learning in dialogue state tracking and intent detection. His ultimate goal is to develop agents which feel human-like in their understanding of speech. In his free time, he enjoy developing simple computer games.
Bordalo holds a BSc degree in CS and is currently a MSc CS student. He has focused his studies in the fields of Machine Learning and Deep Learning. Currently he's researching multimodal models with a focus on Visual Question-Answering, looking to leverage state-of-the-art models for different applications related to Conversational Agents.
Simões is a final year masters student in computer science and engineering. She is focused towards software engineering and her main focus is on requirement analysis and building a flexible conversational agent platform with support for automated testing. She will start her masters' thesis which will be based on these topics.
Valério is currently pursuing a MSc on general purpose vision-language models capable of solving multiple tasks with a single model. He has previously worked with multi-modal models in the conversational domain and has done research on optimizing techniques for computer vision models. His main objective is to reduce the cost of running high-cost applications which require multiple models by replacing them with general purpose models.
João Magalhães - Faculty advisor
Magalhaes is an Associate Professor at the Department of Computer Science of the Universidade NOVA de Lisboa. He holds a Ph.D. degree (2008) in Computer Science from Imperial College London, UK. His research interests cover the different problems of vision and language understanding and multimodal conversational search and AI. He is actively engaged in industry funded R&I projects (e.g. Amazon, Farfetch, BBC, VisionBox). He was the Faculty Advisor of the 2021 Alexa TaskBot Challenge team TWIZ. He was the General Chair of ACM Multimedia 2022 and is one of the organizers of the ACM Workshop in Multimodal Conversational AI.