大规模语言模型增强的中文篇章多维度阅读体验量化研究(Quantitative Research on Multi-dimensional Reading Experience of Chinese Texts Enhanced by Large Language Model)

Jiadai Sun (孙嘉黛), Siyi Tang (汤思怡), Shike Wang (王诗可), Dong Yu (于东), Pengyuan Liu (刘鹏远)


Abstract
“现有的文本分级阅读研究往往从文本可读性的角度出发,以离散的文本难度等级的形式为读者推荐阅读书目。目前,仍缺少一种研究读者在阅读过程中产生的多方面、深层次阅读体验的体系结构。对此,我们调研了读者在阅读中文篇章过程中产生的不同阅读体验,提出了中文篇章多维度阅读体验的量化体系。我们将阅读过程中呈现的连续性的阅读体验归纳为多种类别,并在此基础上构建了中文篇章多维度阅读体验数据集。同时,我们探究了以大规模语言模型为基础的ChatGPT对阅读体验的量化能力,发现其虽具备强大的信息抽取和语义理解能力,在阅读体验的量化上却表现不佳。但我们发现大规模语言模型所蕴含的能力能够以知识蒸馏的方式协助深层属性的量化,基于此,我们实现了大规模语言模型增强的中文篇章多维阅读体验量化模型。模型在各维度阅读体验上的平均F1值达到0.72,高于ChatGPT的Fewshot结果0.48。”
Anthology ID:
2023.ccl-1.32
Volume:
Proceedings of the 22nd Chinese National Conference on Computational Linguistics
Month:
August
Year:
2023
Address:
Harbin, China
Editors:
Maosong Sun, Bing Qin, Xipeng Qiu, Jing Jiang, Xianpei Han
Venue:
CCL
SIG:
Publisher:
Chinese Information Processing Society of China
Note:
Pages:
364–376
Language:
Chinese
URL:
https://aclanthology.org/2023.ccl-1.32
DOI:
Bibkey:
Cite (ACL):
Jiadai Sun, Siyi Tang, Shike Wang, Dong Yu, and Pengyuan Liu. 2023. 大规模语言模型增强的中文篇章多维度阅读体验量化研究(Quantitative Research on Multi-dimensional Reading Experience of Chinese Texts Enhanced by Large Language Model). In Proceedings of the 22nd Chinese National Conference on Computational Linguistics, pages 364–376, Harbin, China. Chinese Information Processing Society of China.
Cite (Informal):
大规模语言模型增强的中文篇章多维度阅读体验量化研究(Quantitative Research on Multi-dimensional Reading Experience of Chinese Texts Enhanced by Large Language Model) (Sun et al., CCL 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.ccl-1.32.pdf