|
25-26 Oct 2024 Rennes (France)
|
|
|
|
Kristopher KyleKristopher Kyle is an Associate Professor of Linguistics and Director of the Learner Corpus Research and Applied Data Science Lab at the University of Oregon. His research interests include natural language processing, corpus linguistics, second language writing, second language assessment, and second language development. Leveraging LLMs in computer assisted language learning: A call for construct validity
In this talk, Kristopher Kyle provides a brief overview of the use of natural language processing in research related to language learning and assessment over the past 50 years, culminating in the advent of large language models (LLMs). He then briefly discusses some recent (mis)uses of LLMs in language learning and assessment research. He argues that while some black-box LLM-based systems can achieve results with reasonable metrics, such systems tend to have particularly weak validity arguments. He then argues for the development and use of LLM-based systems that increase the construct validity of common CALL applications such as automated evaluation and feedback systems and introduce some working examples of these systems.
|
| Online user: 2 | Privacy |
|
