TTLab at AlexandriaX-2026: A Fine-Tuned Surface Tagger for Arabic Machine-Translation Error-Span Detection and Classification

Ali Abusaleh, Bhuvanesh Verma, Alexander Mehler
9/24/2026
cs.CLcs.AIcs.LG

Official code: ENTAILab/arabic-dialectal-mt-error-span-detection (0 stars, last commit 2026-07-09).

Abstract

We present TTLab's submission to the AlexandriaX-2026 Subtask~3 on Arabic MT error span detection and classification. Our system frames the task as token-level classification over surface forms, preserving character offsets to ensure exact alignment with the evaluation metric. To handle severe label imbalance, we employ a focal loss with class weighting and dialect-specific decoding thresholds. Among six Arabic pre-trained encoders, MARBERTv2 achieves the best overall performance of 40.8 and 40.91 on the development and test set, respectively, ranking $\nth{3}$ out of all participating teams. While our system localizes error spans effectively, classification of rare error types remains challenging, highlighting the need for data augmentation for tail categories. The code is available at ${\href{https://github.com/ENTAILab/arabic-dialectal-mt-error-span-detection}{\faGithub~ TTLab at AlexandriaX-2026}$

View on arXivView PDF

Code Implementations(1)

ENTAILab/arabic-dialectal-mt-error-span-detectionOfficial100%

TTLab system for AlexandriaX-2026 Subtask 3: Arabic Dialectal MT Error-Span Detection and Classification

00Jupyter NotebookJul 10, 20263 months ago
arabic-nlperror-detectionllmtranslation

Similar Papers

Unify word-level and span-level tasks: NJUNLP's Participation for the WMT2023 Quality Estimation Shared Task

79% match1 repo122023

A BERT-based Unsupervised Grammatical Error Correction Framework

76% match2023

LED: A Benchmark for Evaluating Layout Error Detection in Document Analysis

76% match2026

Minimum Bayes Risk Decoding for Error Span Detection in Reference-Free Automatic Machine Translation Evaluation

76% match2025

LQM: Linguistically Motivated Multidimensional Quality Metrics for Machine Translation

76% match2026

LogSigma at SemEval-2026 Task 3: Uncertainty-Weighted Multitask Learning for Dimensional Aspect-Based Sentiment Analysis

76% match2026

Cite this paper

@article{abusaleh2026ttlab,
  title  = {TTLab at AlexandriaX-2026: A Fine-Tuned Surface Tagger for Arabic Machine-Translation Error-Span Detection and Classification},
  author = {Ali Abusaleh and Bhuvanesh Verma and Alexander Mehler},
  year   = {2026},
  eprint = {2609.29633},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2609.29633v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2609.29633.svg)](https://codeofpaper.com/paper/2609.29633)

Discussion