Dr Wei Zhao

Biography

Since Jan. 2024, I have been a Lecturer (=Assistant Professor) in Computing Science at the University of Aberdeen, and an Affiliated Lecturer at Heidelberg University where I teach Diachronic Language Models.

Previously, I was a postdoctoral researcher at Heidelberg Institute for Theoretical Studies, by courtesy affiliated with the Institute for Mathematics at Heidelberg University, supported by the Young Marsilius Fellowship (Sep. 2022 - Dec. 2023) and the Klaus Tschira Foundation (Jan. 2022 - Dec. 2023). I earned my PhD at the AIPHES Training Group from Technische Universität Darmstadt, supported by the German National Funding (AIPHES GRK, 2018 - 2021).

My research interests include evaluation for Text Generation (more recently for Generative AI), dynamics of languages, cross-lingual semantic modeling and geometric representation learning. 

• Principles and practices of evaluation [1]: designing metrics that include considerations of semantics [2,3] and discourse coherence [4], carrying out rigorous approaches for system comparison [5], and making the evaluation process accountable [6,7]. 
• Dynamics of languages: detecting and interpreting diachronic changes of languages over a hundred years on the semantic level [8] and syntactic level [9]. 
• Cross-lingual semantic modeling: developing language-agnostic representations through unsupervised approaches [10,11] to democratize language technologies.  
• Geometric representation learning: learning graph embeddings in symmetric space where the space offers rich inductive biases for handling real-world graphs with mixed geometric structures [12], and exploring geometric spaces with good balance between computing efficiency and inductive biases for embedding real-world graphs [13]. 

More recently, has my research expanded to interdisciplinary matters bridging between AI and other disciplines including linguistics and theology where AI is seen as a subject of study, as well as a tool for studying the subjects in other disciplines.