Skip to content

Modular decomposition of AWE into scoring, surface-level feedback, and deep-level feedback

Design · All designs
Evidence · 1 claim (1 for) · 1 study (1 design), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description

The article organizes automated writing evaluation as a modular architecture that divides AWE tasks into three modules: scoring, surface-level feedback, and deep-level feedback. The abstract states that "WrAFT adopts a modular design by dividing the automated writing evaluation (AWE) tasks into scoring, surface-level feedback and deep-level feedback." Deep-level feedback is further split into macro feedback (paragraphing, text length, structure) and micro feedback (context-dependent grammar, clarity, coherence, argumentation, formality), each handled by a separate pipeline. The authors report this decomposition builds on Li and Ng (2024)'s layered proposal for AWE systems and supports specialized models per aspect.

Design Implications

Context

Requirements

  • Separate LLMs or pipelines tailored to each module, with validation schemes appropriate to each (benchmark scores for scoring, human raters for feedback)

Constraints

  • The article notes specialized models handling different aspects can produce better assessments than unified approaches, but the system was developed and tested only on argumentative essays with writing prompts

Target Learners

  • students writing argumentative essays
  • second language writers

Learning Goals

  • essay scoring accuracy
  • surface-level error correction
  • higher-order writing qualities including organization, coherence and argumentation

Claims

Examples

-

Key Sources

  • Adnan Labib, Yixuan Huang, Jiahui Wu, John Maurice Gayed, Zheng Yuan, Qiao Wang. (2026). WrAFT: a Modularized Automated Writing Evaluation System for Argumentative Essays. arXiv preprint. https://arxiv.org/abs/2607.14524
Page metadata
Field Value
type design
id modular-awe-decomposition-framework
title Modular decomposition of AWE into scoring, surface-level feedback, and deep-level feedback
description The article organizes automated writing evaluation as a modular architecture that divides AWE tasks into three modules: scoring, surface-level feedback, and deep-level feedback.
status review
generated by process:settle-candidates on 2026-10-12
trust unverified — no one has confirmed this page's content
sources 1 — adnan-labib-2026

Frontmatter is the page's machine-readable record — see the Schema & Guide.