Graph-Based Multi-Trait Essay Scoring
Educational Testing Service, Presidency University, Fernuniversität Gesamthochschule Hagen, Pohang University of Science and Technology, Soochow University, Capital Normal University, Central China Normal University, University of Stuttgart, Universität Stuttgart, Beijing Normal University, Yeungnam University, University of British Columbia, Dongyang University, Korea Advanced Institute of Science & Technology, Korea Advanced Institute of Science & Technology, University of Maryland, College Park, Guangdong University of Foreign Studies, TSMC, University of Qatar, Peking University, Peking University, Electronics and Telecommunications Research Institute, Westlake University, Kunsan National University, University of Notre Dame, The University of Tokyo, Tokyo Institute of Technology, NTUST, North China University of Technology, Universidad de Chile, Universidad de Chile, New York University, Göteborg University, The Hong Kong University of Science and Technology, Mohamed bin Zayed University of Artificial Intelligence, Sunchon National University, Chung-Ang University, tohmatsu, University of Electro-Communications, Tokyo Institute of Technology, The Hong Kong University of Science and Technology, Hosei University, East China Normal University, Huawei Technologies Ltd., The Robert Gordon University, University of Sheffield, King’s College London, University of London, International Institute of Information Technology, Hyderabad, International Institute of Information Technology Hyderabad, Indian Institute of Technology, Patna, Dhirubhai Ambani Institute Of Information and Communication Technology, University of Southern California, Indian Institute of Technology Bombay, Indian Institute of Technology, Bombay, scnu, Yonsei University, Kyung Hee University, Kyonggi University, Fudan University, alumni.berklee.edu, Indian Institute of Technology, Kanpur, Dhirubhai Ambani Institute Of Information and Communication Technology, University of Surrey, iproov, University of Pittsburgh, Universität Hannover and University of Texas at Dallas
PDF 由论文原始站点提供，PaperCompass 不保存论文文件。DOI 10.18653/v1/2025.emnlp-main.1691 ↗
摘要
While virtually all existing work on Automated Essay Scoring (AES) models an essay as a word sequence, we put forward the novel view that an essay can be modeled as a graph and subsequently propose GAT-AES, a graph-attention network approach to AES. GAT-AES models the interactions among essay traits in a principled manner by (1) representing each essay trait as a trait node in the graph and connecting each pair of trait nodes with directed edges, and (2) allowing neighboring nodes to influence each other by using a convolutional operator to update node representations. Unlike competing approaches, which can only model one-hop dependencies, GAT-AES allows us to easily model multi-hop dependencies. Experimental results demonstrate that GAT-AES achieves the best multi-trait scoring results to date on the ASAP++ dataset. Further analysis shows that GAT-AES outperforms not only alternative graph neural networks but also approaches that use trait-attention mechanisms to model trait dependencies.