System Report for CCL24-Eval Task 7: Multi-Error Modeling and Fluency-Targeted Pre-training for Chinese Essay Evaluation

TL;DR

This paper details a comprehensive system for Chinese essay evaluation, employing multi-error modeling and fluency-focused pre-training, achieving top results in the CCL-2024 challenge.

Contribution

It introduces novel multi-error modeling techniques and fluency-targeted pre-training strategies specifically for Chinese essay evaluation tasks.

Findings

Achieved first place in CCL-2024 fluency evaluation task.

Improved error detection with binary classification models.

Enhanced fluency assessment using back-translation and NSP-based strategies.

Abstract

This system report presents our approaches and results for the Chinese Essay Fluency Evaluation (CEFE) task at CCL-2024. For Track 1, we optimized predictions for challenging fine-grained error types using binary classification models and trained coarse-grained models on the Chinese Learner 4W corpus. In Track 2, we enhanced performance by constructing a pseudo-dataset with multiple error types per sentence. For Track 3, where we achieved first place, we generated fluency-rated pseudo-data via back-translation for pre-training and used an NSP-based strategy with Symmetric Cross Entropy loss to capture context and mitigate long dependencies. Our methods effectively address key challenges in Chinese Essay Fluency Evaluation.