Martin Trapp

Austrian Research Institute for Artificial Intelligence

Papers

Total Citations

H-Index

About

Martin Trapp’s research lies at the intersection of robotics, cognitive science, and computational linguistics, with a focus on how machines can learn language from real-world interaction. His key contributions center on **grounded and cross-situational word learning**—enabling robots to acquire vocabulary not from static datasets, but by observing human actions and objects in dynamic, task-oriented environments. In his most cited work (7 citations), Trapp demonstrated real-time word-object and word-action mapping on a Pepper humanoid robot, allowing it to learn verbs like “take” and “push” through live human demonstration. This work showcases a rare integration of Bayesian inference with embodied robotics, pushing beyond simple label-matching toward genuine referential understanding. His subsequent models (4 citations) further advanced incremental, crossmodal learning from minimal examples—a crucial step for long-term human-robot collaboration. Trapp has also explored strategic dialogue in film (2 citations), analyzing how conversational agents might sustain persuasive goals over extended interactions. Though his citation counts are modest, his work is notable for its ambitious, interdisciplinary approach: bridging formal learning theory with real-time robotic implementation. For students and researchers, Trapp’s research offers a compelling blueprint for building robots that learn language the way humans do—through active, situated experience.