Abstract:
[Purpose/Significance] With the rapid development of Generative Artificial Intelligence (AIGC), large language models have demonstrated powerful language understanding and generation capabilities in general domains. However, they still face many limitations in the field of ancient Chinese processing. To address this challenge, Huazhong University of Science and Technology has developed the "AI Jiusi," a large language model for cognition of ancient Chinese, aiming to enhance the professional capabilities of LLM in knowledge question-answering and comprehension applications related to ancient Chinese. [Method/Process] This paper provides a detailed introduction to the research and development background, dataset construction, model training process, and performance in terms of ancient Chinese language knowledge and linguistic ability of "AI Jiusi." [Results/Conclusions] Based on internal testing feedback, "AI Jiusi" has shown significant advantages in professional question-answering and comprehension application tasks related to ancient Chinese, although there are areas that need improvement. In the future, the team plans to further enhance the text cognition and multimodal application capabilities of "AI Jiusi," optimize user interaction experience, and promote the development of LLMs for ancient Chinese to a higher level, facilitating the transition of ancient Chinese research into the digital and intelligent phase.
-
From:
刘根辉
-
Subject:
Linguistics and Applied Linguistics
>>
Linguistics and Applied Linguistics
Library Science,Information Science
>>
Information Science
-
Contribution:
No Submitted
-
Cite as:
ChinaXiv:202501.00212
(or this version
ChinaXiv:202501.00212V1)
DOI:10.12074/202501.00212
CSTR:32003.36.ChinaXiv.202501.00212
-
TXID:
86fdf660-c1aa-4654-9d91-f7d95fb8d2cc
- Recommended references:
刘金柱,王锦绣,罗捷春,李志芳,袁方,余静静,龚丹,谢雨霏,罗婉滢,郑苏楠,陈旷心,贺心雨,张润哲,夏婉婷,谢佳延,吕佳源,吕萍,余乐妍,郑诗铭,王金柳,刘艺溶,徐君词,张雪晨,冷谦益,杨纯,彭立雪,张曼丽,吴翊嘉,李祎萌,余锁湘,汪靓,刘根辉.AI九思:用大语言模型焕新古汉语之美.中国科学院科技论文预发布平台.[DOI:10.12074/202501.00212]
(Click&Copy)