Keynote speakers

LU Xiaofei, PhD

George C. and Jane G. Greer Professor of Applied Linguistics and Asian Studies

Department of Applied Linguistics

Pennsylvania State University

Keynote Title

Large language model-based approaches to language assessment

Bio

Professor Xiaofei Lu is the George C. and Jane G. Greer Professor of Applied Linguistics and Asian Studies at The Pennsylvania State University. His research interests include corpus linguistics, English for Academic Purposes, second language writing, second language acquisition, and computer-assisted language learning. He serves as an Associate Editor for the Journal of Second Language Writing. He has published over 110 peer-reviewed articles in leading journals. Since 2022, he has been recognized on Stanford University’s annual list of the Top 2% Most-Cited Scientists Worldwide in the subfield of Languages & Linguistics for both single-year and career-long impact.

Abstract

Embedding-based approaches to writing assessment offer an alternative to both traditional feature-engineering methods and direct large language model scoring. This talk first defines what an embedding-based approach is and traces developments in text representation methods from static word embeddings to contextualized token embeddings and from word-level to sentence- and document-level embeddings. It then discusses several recurring questions in recent research: whether embeddings improve score prediction, what aspects of writing quality and proficiency they capture, how they can be used to measure essay-prompt relevance, how well they generalize across prompts and languages, and whether they should replace or complement corpus-derived linguistic features. Particular attention is given to the emerging methods scholars have used to address these questions and their implications for new assessment practices. The talk concludes with a research agenda emphasizing construct validity, interpretability, multilingual and multimodal applications, and closer integration between corpus-informed linguistic measures and large language model representations.

Disclaimer

Site Map