Embodied Conversational Agents: Learned Lessons and opportunities

• TUESDAY 2-December-2022, 16:00 A 18:00
• AULA 104  | Sala en Teams: http://bit.ly/3U1blia
• ESCUELA DE INGENIERÍA INFORMÁTIVA DE VALLADOLID

SUMMARY

This seminar will cover past experiences we had in developing ECAs for various case studies, from investigating the role of non-verbal behaviour in cultural heritage presentation to movie recommendation and speech therapy tests administration. We will summarise the lessons learned during these experiences and how they influenced the development of FANTASIA which, other than being a tool, aims at representing a working method that combines the expertise of researchers from the humanities areas and the one of researchers coming from the technological areas. Through collaboration between these profiles, we present a research program to develop machines that go beyond mechanical service providers and support users in reflection, entertainment, and self-improvement.

SPEAKERS

Doctor Antonio ORIGLIA

• https://scholar.google.it/citations?user=C_WbH78AAAAJ&hl=it
• https://www.linkedin.com/in/antoniooriglia

Antonio ORIGILIA is an assistant professor at University of Naples “Federico II” from 2021, where he was postdoctoral researcher from 2017 to 2020. During his PhD, he concentrated on Affective Computing, focusing on emotional speech analysis with robotics applications. His work consisted of developing acoustic features extraction algorithms to perform continuous emotion recognition on the valence and arousal axes using machine learning approaches. His PhD thesis was awarded an honorable mention in the national "Antonio Zampolli" prize for the best Italian PhD theses in Computational Linguistics. After his PhD, he concentrated on Human-Computer Interaction topics, mainly focusing on applications for Cultural Heritage. His job has consisted of working with art historians and linguists to develop technological applications aimed at promoting cultural contents using new technologies. During this work, he acquired competence in 3D graphics, animation and he had significant experience using the Unreal Engine 4. He also acquired competence in designing dialogue systems based on the use of probabilistic graphical models and graph databases. During his work at the University of Padova, he applied these competences to design and implement an interactive system for speech therapy applications. This system was designed to use virtual characters, both 3D avatars and humanoid robots, to administer adaptive acoustic discrimination tests to young children (5 years old). The system would, then, provide speech therapists with detailed information about the children's performance. During and after his PhD, he supervised bachelor and master students working on the topics of Speech Processing and Human-Computer Interaction. He also co-supervised PhD students working on the same topics and held seminar series for students enrolled in the Computer Science Bachelor and Master degree courses at the University of Naples "Federico II". He has teaching experience in the fields of Speech and Language Processing and in the field of Human - Computer Interaction, concentrating specifically on probabilistic dialogue systems and their use in advanced applications developed using UE4. He has participated in writing multiple national, regional and European projects, gaining significant experience in managing research consortia involving both industrial and academic partners.

Doctor Loredana Schettino

Loredana Schettino is a postdoc researcher at theUniversity of Naples “Federico II”, where she is also adjunct professor ofGeneral Linguistics. She obtained her Ph.D. in Linguistics from the Universityof Salerno, discussing a thesis entitled “The Role of Disfluencies in ItalianDiscourse. Modelling and Speech Synthesis Applications”. The project has focused on the ways phenomena, such as silent pauses, prolongations, fillers contribute to communication as convenient tools at speakers’ hand to gain valuable time for speech planning and, at the same time, for listeners’ information processing. The work shows that neural speech synthesis may be considered as a useful tool for evaluating linguistic models non-verbal pausing behaviour. Concurrently, it highlights the relevance of such models for the development of natural sounding and communicative efficient speech technological applications.Her current project (since October 2022) concerns the analysis of prominence patterns and pausing behaviours in Italian informative speech aimed at deriving linguistic models to be implemented in the generation of synthetic voices supporting cultural content presentation byVirtual Humans.Her main research interests concern the study of phonetic and functional features of non-verbal linguistic phenomena in Italianspeech, also in contrastive perspective, and their contribution to communication,also taking into account their relationship with other non-verbal behaviours such as gestures. Further research activities concern investigations in contrastive perspective on the linguistic factors that affect syllabic phonetic variation and reduction phenomena in speech.Also, she has been involved in pragmatic studies in contrastive perspective related to the conversational strategies adopted by speakers in dyadic interactions.