ECNU Review of Education unveils the evolution of LLMs in intelligent Chinese composition tutoring

Writing is a fundamental aspect of language learning, but the process of grading compositions often proves time-consuming and inconsistent. Teachers struggle with varying grading standards, and the sheer amount of time spent on grading limits their ability to provide meaningful, interactive feedback. These challenges are particularly prominent in Chinese language education, where automated essay scoring has not been as extensively developed as it has in English-language contexts.

To address these challenges, a team of researchers comprising Chanjin Zheng and Wei Xia from East China Normal University, along with Shaoguang Mao and Yan Xia from Microsoft Research Asia, has developed the ELion Intelligent Chinese Composition Tutoring System, an AI-powered solution designed to enhance essay evaluation. Their study, published online in ECNU Review of Education on February 4, 2025, explores the system’s evolution and its impact on Chinese language education.

"We've witnessed a remarkable transformation in how AI can support Chinese language education," add Zheng et al. "The evolution from BERT to the integration of ChatGPT and beyond demonstrates the potential of combining different AI technologies to create comprehensive educational solutions."

In the study, the researchers explain how the ELion system operates by evaluating student essays across four key dimensions: topic comprehension, content quality, expression (including fluency, logic, and grammar), and handwriting. Initially utilizing the Bidirectional Encoder Representations from Transformers (BERT) model, the system analyzed the linguistic features of student essays, providing detailed, objective assessments. This model was further enhanced with the integration of ChatGPT, enabling the system to offer more nuanced and personalized feedback, which in turn fosters student creativity and interaction.

The integration of ChatGPT with BERT has significantly improved how feedback is delivered. The system not only generates comprehensive reviews of students’ writing but also fosters a more interactive dialogue between the AI and students, encouraging continuous improvement of their writing skills.

The study highlights that the ELion system has been instrumental in addressing the longstanding issue of teachers' excessive grading workloads, particularly in Chinese primary and secondary schools. Since its launch in the spring of 2021, the system has expanded rapidly, with 251 schools adopting the technology across China by June 2024. With over 15,000 students and 560 teachers using the platform, the system has facilitated the analysis of nearly 50,000 compositions.

One teacher from a primary school in Shanghai emphasized the value of the ELion system in the classroom: "ELion has liberated Chinese language teachers from the tedious task of essay correction, allowing them to focus more on interactive learning and creativity development." The system has also helped alleviate the burden on teachers, freeing up time for more focused, creative teaching.

Despite its success, the researchers acknowledge that the widespread implementation of AI in education presents challenges. While technologies like ELion have shown strong potential to improve grading efficiency and deliver personalized feedback, it remains essential to integrate these tools thoughtfully into classroom settings. The real value of AI lies in its ability to support—not replace—the human aspects of teaching.

The study also discusses the importance of evolving AI systems to adapt to diverse educational needs. The integration of BERT and ChatGPT into the ELion system serves as a model for how AI can be tailored to support various educational contexts. The researchers highlight that the ability to combine traditional and cutting-edge technologies is key to ensuring that AI-driven tools can offer meaningful contributions to the educational landscape.

As the ELion system continues to expand its reach and refine its capabilities, the researchers emphasize that AI in education must remain flexible and adaptive to the needs of students and teachers alike. They point out that by utilizing a combination of AI tools, a more personalized, efficient, and engaging learning environment can be created, one that prepares students for the challenges of the future.

This study offers valuable insights into the potential of AI in transforming education, highlighting the role of advanced technologies in providing personalized learning experiences and alleviating the workload of teachers. The ongoing evolution of the ELion system sets a precedent for future innovations in AI-powered educational tools.

***

Reference

DOI: 10.1177/20965311251315205