MSc Thesis Proposals
This is a list of MSc thesis proposals related to my projects. The proposed projects are rough ideas, and we can discuss more precise (or alternative) research questions together. Please if you are interested!
Modelling turn-taking in conversational systems
One of the most fundamental aspects of spoken interaction is the taking of turns. Many studies have investigated the mechanisms by which speakers manage this coordination, for example by identifying acoustic and linguistic signals at the end of the turn that are associated with turn shifts. Compared to humans, current conversational systems are very bad at turn-taking and often tend to interrupt the user or give sluggish responses. In our research, we develop models that can predict human turn-taking and test these models in conversational systems.
Potential MSc project: Apply and adapt our data-driven models of turn-taking to a conversational system and evaluate it.
Social robots as hosts on self-driving buses
Together with ITRL at KTH, we are exploring the use of social robots as hosts on board self-driving buses. Currently, we are doing experiments and data-collections where the Furhat robot is placed on a self-driving bus and interacts with passengers.
Potential MSc project: Do an experiment where you compare a robot that is more proactive (initiates interactions with the passengers) with one that is more passive.
Social robots for language learning
We are part of the larger EU project e-LADDA , which investigates digital tools for language learning. At KTH, we are investigating the use of social robots (such as Furhat) for language learning. We are especially interested in exploring incidental learning, i.e., where you do not necessarily interact with the robot primarily to learn language, but where that is a secondary outcome.
Potential MSc project: Set up a human-robot language learning experiment where different pedagogical strategies are compared.
Visual grounding of language in dialogue
A fundamental requirement for any intelligent system that is situated in a physical environment – and that should also be able to reason symbolically about this environment or communicate about it using symbolic language – is that it can understand the relationship between these symbols, the objects or phenomena they denote, and their properties. In this project, we develop deep learning methods for connecting vision, language and dialog. Such models can then be used to allow robots to communicate with humans about objects in the shared environment.
Potential MSc project: Explore how the recently developed CLIP model by OpenAI can be used in an interactive human-robot interactive setting.
Social robots as virtual patients
A potential application area for social robots is to simulate patients, so that medical students can practice on what it is like to talk to those patients. In a collaboration with Karolinska Institutet (KI), we want to develop a prototype platform for such virtual patients in the domain of rheumatic problems.
Potential MSc project: Develop and evaluate a prototype of a virtual patient platform using the Furhat robot.