Research on Detecting AI-generated Chinese Academic Text
Widespread application of the AIGC technology such as ChatGPT provides great convenience for AI automatically generating text content including papers and reports,urging the administrative de-partments in fields of publication and education to perform AIGC detection on submitted texts utilizing intelligent technology.In this paper,text detection methods for Chinese academic paper texts generated by the AI continuation and the AI expansion respectively are explored and experimented,which are two familiar human-computer collaboration methods.A human-computer interaction-based approach for gen-erating and detecting multidisciplinary academic paper text data is designed,where four detection data-sets generated by different large language models are adopted.Detection is performed by using the tool AIGC-CHECK to assess the effectiveness of the approach in actual scenarios.Detection results indicate that AI annotation accuracy using the tool AIGC-CHECK optimally reaches 98%.Compared with detec-ting text of Chinese academic papers completely generated by AI,our approach could still achieve high accuracy for the AI continuation and the AI expansion which demand greater difficulty and finer detec-tion granularity,demonstrating the feasibility of applying such methods to assist in detecting AI-genera-ted academic papers.
Chinese academic paperAIGC detectionAI expansion detectionAI continuation detec-tion