Large Language Model-Assisted Move Structure Recognition in Abstracts of English Academic Articles from Chemistry Journals
YANG Chuan-Ming1**, KONG Xiang-Rui1, GUO Yun-Jie2, LIU Yu-Xuan1
1. College of Arts and Science, Northeast Agricultural University, Harbin 150030, China; 2. School of Marxism, Jingmen Technical College of General Aviation, Jingmen 448001, China
Abstract The abstracts of English academic articles typically consist of several move structures with explicit functions.Move structure analysis has been a key concern in the academic English writing instruction in various disciplines,however,the analysis has long been constrained by labor-intensive and time-consuming manual recognition.Although large language models(LLMs) present new opportunities for the automated move structure recognition,the research on LLMs in the abstracts of English academic articles from chemical journals is remain insufficient.Therefore,the study compares the performance of DeepSeek-V3.2,GPT-5.1,Gemini 2.5 Pro,and Grok 4.1 Fast in recognizing the moves in abstracts of academic articles from chemistry SCI journals.The results show that DeepSeek-V3.2 and Gemini 2.5 Pro have complementary strengths:DeepSeek-V3.2 excels in precise recognizing,while Gemini 2.5 Pro performs better regarding comprehensive recognizing.In comparison,although GPT-5.1 and Grok 4.1 Fast do not exhibit obvious advantages,they still achieve relatively robust recognition capabilities.The results demonstrate that LLMs have high recognition effectiveness,and reliability,facilitating the automation of move structurerecognition.The findings can provide practical support for Human-AI collaboration instruction of academic English writing instruction in the field of chemistry.
YANG Chuan-Ming, KONG Xiang-Rui, GUO Yun-Jie, LIU Yu-Xuan. Large Language Model-Assisted Move Structure Recognition in Abstracts of English Academic Articles from Chemistry Journals[J]. Chinese Journal of Chemical Education, 2026, 47(10): 92-100.
