Abstract
“Chinese+” represents a pivotal evolution in Chinese for Specific Purposes (CSP), emphasizing practical language application in specialized fields. As international demand for vocational and technical Chinese grows, teachers face the dual challenge of linguistic instruction and domain-specific expertise. This paper explores the integration of Generative AI (GenAI) in constructing “Chinese+” teaching resources, specifically for medical Chinese. Through a comparative analysis of human-authored textbooks and AI-generated content, we identify a "difficulty gap" and propose a "70/30 productivity rule" for future resource development.
Introduction
International Chinese education is increasingly guided by practical needs, highlighting the effectiveness of specialized instruction in fields like medicine, engineering, and trade. However, most language instructors lack the deep domain knowledge required to author authentic, technically accurate textbooks. Generative AI offers a potential bridge for this knowledge gap, yet its pedagogical alignment—specifically regarding linguistic difficulty and terminology density—remains under-evaluated.
Methodology
To test the feasibility of AI-assisted textbook writing, a comparative study was conducted using four prominent Large Language Models (LLMs): DeepSeek, Zhipu Qingyan, Wenxiaoyan, and Doubao.
The researchers developed a structured "training paradigm" (prompt engineering) targeting the Practical Medical Chinese (Basic Edition) curriculum. The models generated over 30,000 characters of content across 10 medical topics, including 20 dialogues and various technical short essays. These outputs were then graded against the International Chinese Education Chinese Proficiency Grading Standards to measure alignment with HSK levels 1-6.
The "Difficulty Inversion" Phenomenon
The study revealed a significant discrepancy between AI-generated content and traditional pedagogical standards, termed "difficulty inversion."
1. Conversational Complexity
AI-generated dialogues tended to be more complex than their textbook counterparts. While the benchmark textbook focused on HSK Levels 2 and 3, the AI outputs frequently reached Levels 3 to 5. This was largely due to the "Colloquialism Trap": models often inserted unnecessary filler phrases and modal particles to sound more natural, which inadvertently increased the linguistic load for beginner students.
2. Technical Essay Simplification
Conversely, when generating specialized short texts (essays), the AI models were often too simple. While the benchmark textbook utilized dense, professional terminology reaching Levels 6–9, the AI-generated essays often fell into Levels 4 and 5. The models tended to "smooth over" technical jargon with common language, failing to meet the high terminology density required for advanced medical Chinese.
The 70/30 Productivity Rule
Despite these mismatches, the AI-generated texts were found to be logically sound and factually consistent with medical topics. The research suggests a 70/30 Productivity Rule: Generative AI can handle approximately 70% of the initial workload, including data collection, structural drafting, and basic content generation.
However, the remaining 30%—the "human element"—is critical. Human educators must perform the following tasks:
- Calibration: Adjusting the HSK level to ensure the text matches the students' proficiency.
- Pruning: Removing redundant colloquialisms or "classroom language" generated by the AI.
- Terminology Infusion: Manually increasing the density of professional terms in technical essays to meet advanced standards.
Conclusion and Future Directions
Generative AI is a viable auxiliary tool for "Chinese+" resource construction, particularly for teachers without professional backgrounds in specialized fields. To optimize this technology, future efforts should focus on developing textbook-specific local corpora for model fine-tuning and improving the digital literacy of language teachers. By mastering prompt engineering, educators can better guide AI to produce level-appropriate, domain-specific content that supports the sustainable development of international Chinese education.
Summary updated