Scholarlink

<p>The emergence of generative artificial intelligence (GAI) has introduced significant changes across multiple domains, including medicine. Understanding its impact on language use, particularly in medical abstracts, is essential for maintaining clarity, precision, and objectivity. This study examines changes in verb usage in English-language medical abstracts before and after the widespread adoption of GAI. Specifically, it investigates how verb usage patterns have changed following the emergence of GAI. A large-scale text mining approach was applied to analyze verb usage in 157,720 medical abstracts retrieved from the Web of Science. The study period was segmented into three distinct phases: a baseline period (November 2018–April 2019), chosen to represent a pre-GAI landscape prior to widespread awareness or use of generative tools; a pre-GAI period (May 2021–October 2021), and a post-GAI period (December 2023–May 2024), following the public release and institutional uptake of models like ChatGPT. Verbs were categorized into lexical and auxiliary groups, with frequency normalization to account for variations in abstract count across time periods. Statistical comparisons were conducted using paired <i>t</i>-tests and chi-squared tests to determine significant shifts in overall and individual verb usage, respectively. The change in lexical verb usage from the pre-GAI to the post-GAI period was statistically different from the change observed from the baseline to the pre-GAI period (<i>P</i> = 0.003). Notably, certain lexical verbs associated with deeper analytical narratives saw substantial increases, including ‘delve’ (+ 2665%, <i>P</i> &lt; 0.001), ‘underscore’ (+ 1182%, <i>P</i> &lt; 0.001), and ‘showcase’ (+ 689%, <i>P</i> &lt; 0.001). Conversely, lexical verbs such as ‘reserve’ (− 64%, <i>P</i> &lt; 0.001), ‘fight’ (− 57%, <i>P</i> &lt; 0.001), and ‘transplant’ (− 57%, <i>P</i> &lt; 0.001) exhibited declines. Commonly used verbs, including ‘use’ (+ 5%, <i>P</i> &lt; 0.001) and ‘include’ (+ 16%, <i>P</i> &lt; 0.001), demonstrated only minor fluctuations. Auxiliary verbs showed non-significant trends except for ‘be’ (− 3%, <i>P</i> = 0.006). The findings suggest that the emergence of GAI has coincided with shifts in the linguistic structure of medical abstracts, particularly in lexical verb usage patterns. While these changes may partially reflect evolving writing styles, they also raise questions about the extent to which AI-generated text is influencing scientific discourse.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Transforming world and word: a longitudinal analysis of verb usage within English abstracts from medical articles

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月18日

查看全文
  • Takanobu Hirosawa,
  • Yukinori Harada,
  • Taro Shimizu

摘要

The emergence of generative artificial intelligence (GAI) has introduced significant changes across multiple domains, including medicine. Understanding its impact on language use, particularly in medical abstracts, is essential for maintaining clarity, precision, and objectivity. This study examines changes in verb usage in English-language medical abstracts before and after the widespread adoption of GAI. Specifically, it investigates how verb usage patterns have changed following the emergence of GAI. A large-scale text mining approach was applied to analyze verb usage in 157,720 medical abstracts retrieved from the Web of Science. The study period was segmented into three distinct phases: a baseline period (November 2018–April 2019), chosen to represent a pre-GAI landscape prior to widespread awareness or use of generative tools; a pre-GAI period (May 2021–October 2021), and a post-GAI period (December 2023–May 2024), following the public release and institutional uptake of models like ChatGPT. Verbs were categorized into lexical and auxiliary groups, with frequency normalization to account for variations in abstract count across time periods. Statistical comparisons were conducted using paired t-tests and chi-squared tests to determine significant shifts in overall and individual verb usage, respectively. The change in lexical verb usage from the pre-GAI to the post-GAI period was statistically different from the change observed from the baseline to the pre-GAI period (P = 0.003). Notably, certain lexical verbs associated with deeper analytical narratives saw substantial increases, including ‘delve’ (+ 2665%, P < 0.001), ‘underscore’ (+ 1182%, P < 0.001), and ‘showcase’ (+ 689%, P < 0.001). Conversely, lexical verbs such as ‘reserve’ (− 64%, P < 0.001), ‘fight’ (− 57%, P < 0.001), and ‘transplant’ (− 57%, P < 0.001) exhibited declines. Commonly used verbs, including ‘use’ (+ 5%, P < 0.001) and ‘include’ (+ 16%, P < 0.001), demonstrated only minor fluctuations. Auxiliary verbs showed non-significant trends except for ‘be’ (− 3%, P = 0.006). The findings suggest that the emergence of GAI has coincided with shifts in the linguistic structure of medical abstracts, particularly in lexical verb usage patterns. While these changes may partially reflect evolving writing styles, they also raise questions about the extent to which AI-generated text is influencing scientific discourse.