Under the premise of ensuring accuracy, efficiently identifying focal information in sentences via test methods is fundamental for effective communication, especially for the elderly who struggle with extracting core reading content. “Shi… de”, “shi”, and “de shi” are common focus markers in Chinese, indicating contrastive foci with exclusivity and prominence that are vital for the elderly’s information decoding. In the big data era, leveraging artificial intelligence (AI) for focus recognition offers a promising way to optimize age-appropriate reading assistance systems. This study tests the performance differences of AI models (ChatGPT, DeepSeek, Qwen) in recognizing these focus-marked sentences before and after machine learning. A dataset of 1,100 expert-annotated samples was used, with fine-tuned models via the LoRa method and Unsloth 2.3 framework for memory optimization. Baseline tests showed AI accuracy lagged far behind manual recognition, with varying performance across marker types. Post fine-tuning, significant improvements were achieved, verifying machine learning’s potential to enhance AI focus recognition. This research provides a feasible pathway for upgrading age-appropriate readers, helping the elderly better grasp key reading information.
References
[1] Yuan, F. (2022) Facilitative effects of form-focused tasks on teaching and learning Chinese aspect marker le. Pedagogical Grammar and Grammar Pedagogy in Chinese as a Second Language, 152-166.
[2] Cruschina, S. (2021) The greater the contrast, the greater the potential: on the effects of focus in syntax. Glossa, 6(1), 3.
[3] Tang, P., Yuen, I., Demuth, K., Xu Rattanasone, N. (2023) The acquisition of contrastive focus during online sentence-comprehension by children learning Mandarin Chinese. Developmental Psychology, 59(5), 845.
[4] Lin, Y. (2018) The role of educational factors in the development of lexical splits. Asia-Pacific Language Variation, 4(1), 36-72.
[5] Spalek, K., Oganian, Y. (2019) The neurocognitive signature of focus alternatives. Brain and Language, 194, 98-108.
[6] Shyu, S. I. (2018) The scope of even: Evidence from Mandarin Chinese. Language and Linguistics, 19(1), 156-195.
[7] Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Chen, W. (2022) Lora: Low-rank adaptation of large language models. ICLR, 1(2), 3.
[8] Ding, N., Qin, Y., Yang, G., Wei, F., Yang, Z., Su, Y., Sun, M. (2023) Parameter-efficient fine-tuning of large-scale pre-trained language models. Nature Machine Intelligence, 5(3), 220-235.
[9] Li, Y., Cheng, H., Qin, Q. (2025) Evaluations and improvement methods of deep learning ability in blended learning. International Journal of e-Collaboration (IJeC), 21(1), 1-17.
[10] Anisuzzaman, D. M., Malins, J. G., Friedman, P. A., Attia, Z. I. (2025) Fine-tuning large language models for specialized use cases. Mayo Clinic Proceedings: Digital Health, 3(1), 100184.
Share and Cite
Du, K. (2025) Research on Optimization of Aging Reading Assistance System Based on Artificial Intelligence Focus Recognition. Scientific Research Bulletin, 2(5), 28-34. https://doi.org/10.71052/srb2024/DYRL8667