Abstract:
Language acts as the medium for information transfer in human society, through which meaning is continuously generated, conveyed, and interpreted. The evolution of large language models (LLMs) put it more challenging and provide insights that meanings, rather than structures, might primarily organize and convey the information. We aim to explore to what extent a LLM might share similarities with human’s behaviours in realistic natural discourse reading. We firstly abstracted LLM-based semantic metrics to quantitively model the semantic representation in a dynamic and multilevel nature. Eye-tracking data were recorded when healthy adult, typically developed children and children with developmental dyslexia read Chinese paragraphs. The LLM-based semantic metrics were found significantly correlated to eye-movement features. In dyslexic individuals, this approach further captured discourse-level reading processing specificity and demonstrated diverse linking patterns. These links can then assist LLMs to generate texts tailored for those with reading disorders: we explored a promising approach to optimize reading texts and generate tailored training material. It would formulate effective personalized intervention strategies, ultimately addressing the special educational needs of children with developmental dyslexia with greater efficacy.

Biography:

Lu Shuo, Ph.D. in Linguistics, is a Distinguished Professor and doctoral supervisor at the College of International Studies, and Director of the Neurolinguistics Laboratory at Shenzhen University. She received dual bachelor’s degrees in Literature and Economics from Peking University in 2005, completed her master’s degree at Peking University in 2008, and earned his Ph.D. from City University of Hong Kong in 2011. She also serves as a Research Fellow in the Department of Linguistics and Modern Languages at the Chinese University of Hong Kong.

Professor Lu is the Principal Investigator of the major National Social Science Fund project “Research on corpus construction and language evaluation of aphasia in China” and has been recognized as a High-level Talent in Language Application by the Ministry of Education and as a Young Talent of Guangdong Province (Pearl River Scholar). She has led multiple projects funded by the National Social Science Fund, the Ministry of Education, and Guangzhou Science and Technology Program, and is the director of the Guangdong High-tech Language Service Science Popularization Demonstration Base. She also leads the “Leading Scholar” Innovation Team in Humanities and Social Sciences at Shenzhen University. She has published 30 research articles in SCI-indexed journals such as IEEE Transactions on Medical Imaging, NeuroImage, Learning and Instruction, Cerebral Cortex, and Journal of Speech, Language, and Hearing Research (JSLHR), along with 12 papers in CSSCI-indexed journals including Zhongguo Yuwen. She holds more than 10 national patents.