About me

Hi! I’m a PhD student in the CDT in Natural language processing at the University of Edinburgh, where I work with Sharon Goldwater and Hao Tang, and collaborate with Naomi Feldman.

In my research, I study how self-supervised models of speech encode information about speakers or about context (neighboring phones). More generally, I want to understand how humans and models acquire and process language. Specifically, I am interested in the representational geometry of different types of information, and whether certain geometrical properties (1) resemble those of humans’ mental representations or (2) reflect better downstream performance of the model.

More generally, I am interested in how humans and machines acquire and process language. From a cognitive science perspective, I view machine learning models as a powerful instrument to experiment with theories of cognitive processes. On the more practical side, I am interested in interpretability for models of speech, which is a relatively understudied area as compared to interpretability for NLP, using theories from linguistics and cognitive science.

I’m doing quite a bit of teaching support this year, if you’d like to give me anonymous feedback, click here. Alternatively, email me (oli.liu) here: ed.ac.uk to talk about my research or PhD application or anything about your course :)

Last but not least, when in doubt, wiggle it out✨.

News

September 2024 - This past summer, I’ve been fortunate to attend HILS in Nijmegen (poster), Cogsci in Rotterdam (talk), and Interspeech in Kos (poster). Now I’m ready to stay away from making posters and writing talks, and get some work done ✍️💻📖.

June 2024 - I gave a talk at the University of Pompeu Fabra and Barcelona Supercomputing centre on June 26 and 28.

May 2024 - Our CogSci 2024 paper won the Computational Modeling Prize in Perception & Action!