Monday, 2025-02-17

1. Renata Geld is Associate Professor and Founding Director of the Center for Cognitive Science and Director of Doctoral Program in FL education, Faculty of Humanities and Social Sciences, University of Zagreb. She does research in cognitive semantics, cognitive science and creativity & innovation, creativity and conceptual integration, visual grammar and multimodal meaning construal, cognitive strategies and conceptual integration in L1 and L2.
2. Dario Bojanjac Associate Professor at the University of Zagreb, Faculty of Electrical Engineering and Computing.

Title: Decoding Meaning and Form: Dialogues Between and About Human and Artificial Minds Across Classrooms and Disciplines

Abstract: Creating authentic learning environments, enhanced by real-world engagement or immersive, technology-driven simulations, means creating a space of active participation, teamwork, diverse perspectives, “learning from the future” (Pechl, 2023), “what-ifs,” and conceptual blends (Fauconnier & Turner, 2002), reflection, and the meaningful application of knowledge. These environments have the potential to prepare students for the VUCA (volatile, unpredictable, complex, ambiguous) world, where human and artificial minds are increasingly intertwined. This unavoidable entanglement calls for collaboration across disciplines, especially a meaningful conversation between the humanities and STEM. In my effort to create an authentic new learning environment, my students have produced challenging questions—many of which have emerged from their discussions on language and construction of meaning, meaning and form, intentionality and understanding, humans and machines, cognition and computation: 

• Can meaning exist independent of form, or is the structure of language and thought inseparable from how we understand the world?
• Do we always want systems that replicate the human division of the world, or might AI offer new ways of interpreting meaning that we haven't yet imagined?
• If a machine can “understand” meaning, does it truly grasp human experience, or is it just decoding patterns we’ve created?
• What role does imagination play in human cognition, and could machines, in some way, enhance or even replicate this quality?
• How might we rethink our approaches to learning if we accept that both human cognition and computational models rely on abstract forms but interpret them very differently?
• Is there a limit to how well computational models can simulate human cognition, or are we getting closer to truly artificial understanding?
• How do we explain language acquisition: Is it driven by innate mechanisms, or can it be fully understood through interaction and learning from the environment, as in current AI models?
• Can computational models of language acquisition (like connectionist models) truly replicate the way humans acquire language, or must we look beyond data-driven approaches to capture the complexity of human learning?
• Are there limits to how well artificial systems can model second language acquisition (SLA)?

References

1. Fauconnier, G., & Turner, M. (2002). The way we think: Conceptual blending and the mind's hidden complexities. Basic Books.
   
2. Peschl, M.F. (2023). "Learning from the future as a novel paradigm for integrating organizational learning and innovation", The Learning Organization, Vol. 30 No. 1, pp. 6-22. https://doi.org/10.1108/TLO-01-2021-0018

Past Colloquia

Tiago Torrent is a Cognitive Linguist working on Natural Language Understanding within the framework of Frame Semantics and Construction Grammar. Torrent is head of the FrameNet Brasil Computational Linguistics Lab, PI of ReINVenTA, and Professor of the Graduate Program in Linguistics at the Federal University of Juiz de Fora, Brazil. He is Research Productivity Grantee of the Brazilian National Council for Scientific and Technological Development (CNPq), and Guest Professor at the Department of Swedish, Multilingualism and Language Technology at the University of Gothenburg.

Title: Frame-Based Models of Linguistic Cognition for Social Impact: Insights from the Data to Stop GBV Initiative

View the Recording

Abstract: This talk explores the potential of domain-specific framenets in the semantic analysis of open-text fields in e-medical records, aiming to identify early patterns of gender-based violence (GBV) before escalation. I compare the performance of a frame-based AI model with two alternatives: one relying solely on demographic data and another leveraging a large language model (LLM). The results highlight the superior performance of the frame-based approach, emphasizing the value of human-curated, perspectivized models of linguistic cognition. These findings underline the critical role of interpretable and responsible AI in addressing pressing societal challenges, such as GBV prevention.

Marcelo Viridiano, PhD, is a postdoctoral researcher at Case Western Reserve University.

Title: Multimodal Research in Vision and Language:
     Emerging Trends and the Role of Frame Semantics.

View the Recording

Abstract: Deep learning and its applications have driven transformative research and development across a wide range of real-world data modalities. Recently, this progress has increased the interest in the intersection of vision and language—a rapidly growing field with numerous applications. In this talk, we aim to: Provide a brief review of recent VisLang research, highlighting the latest trends in visual and language modalities. Explore key applications and their approaches to addressing challenges in semantic perception, content analysis, and content generation. Discuss ongoing research by FrameNet Brasil and examine the potential contributions of Frame Semantics to advancing the field.

Monday, 2025-01-27

Fred Belcavello, PhD, is a postdoctoral researcher at Case Western Reserve University, a Researcher at FrameNet Brasil Lab, and a Professor at UniAcademia - Centro Universitário Academia.

Title: Frame Semantics, FrameNet, and the Multimodal Shift

View the Recording

Abstract: In this talk we will present the general idea of semantic frames and how they are represented in its computational implementation, FrameNet, as a language model. We go through FrameNet fundamentals - frames, frame elements, frame to frame relations - and present how it started as a lexicographic project and how we shifted it to a multimodal resource for Natural Language Processing projects.