Morocco

This paper addresses key gaps in assessing Dialectal Arabic to Modern Standard Arabic (DA-MSA) translation quality, where standard automatic metrics and general-purpose human evaluation methods fail to capture dialect-specific errors. We introduce Ara-HOPE, a human-centric post-editing evaluation framework comprising a five-category error taxonomy and a decision-tree annotation protocol designed for consistent use by minimally trained native dialect speakers. Using Ara-HOPE, we comparatively evaluate three mainstream systems (Jais, GPT-3.5, and NLLB-200) to quantify error patterns, pinpoint systematic weaknesses, and provide actionable guidance for improving dialect-aware MT. Our analysis shows clear performance differences across systems and highlights dialect-specific terminology handling and semantic preservation as the most persistent challenges in DA-MSA translation.

EACL 2026

Ara-HOPE: Human-Centric Post-Editing Evaluation for Dialectal Arabic to Modern Standard Arabic Translation

machine translation machine translation evaluation human-evaluation

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

ACL 2024

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

workshop paper

#### *Message from the General Chair, Aline Villavicencio*
I’m delighted and honoured to welcome you to the 19th Conference of the European Chapter of the Association for Computational Linguistics (EACL 2026), taking place in the beautiful city of Rabat, in Morocco, in March 24-29, 2026. EACL is the flagship European conference of the Association and EACL 2026 proudly continues our field’s tradition of excellence in scholarship, innovation, and inclusivity. I am deeply grateful to the many volunteers whose dedication, generosity, and tireless efforts have made this conference possible.
For the first time EACL is being hosted in the African continent. This is an important milestone for our community, and we are grateful to our Moroccan hosts for enabling this historic moment by bringing this edition of EACL to Rabat. We are also delighted that the Second Arabic NLP School is co-located with EACL. We hope attendees enjoy this wonderful opportunity to strengthen ties with the Computational Linguistics communities across the African continent. *[Read full message](https://drive.google.com/file/d/14NlmHvwM6fPJuMmOvVh7K0vtQbEyv3SZ/view?usp=sharing)* 

<html><button style="display: inline-flex; align-items: center; justify-content: center; white-space: nowrap; border-radius: 9999px; font-weight: bold; background: #7c3aed; color: white; font-family: 'Space Grotesk', sans-serif; height: 40px; font-size: 16px; padding: 0 20px; border: none; cursor: pointer" onclick="window.open('https://underline.io/events/522/reception','_blank')">Go to Workshops and Tutorials Program</button></html>

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to EACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.eacl.org/registration/) first.

**Online Registration Form**: https://acl.swoogo.com/eacl2026

Registration Required

Welcome to the 19th Conference of the European Chapter of the Association for Computational Linguistics (EACL). EACL 2026 will be held in Rabat, Morocco, from March 24–29, 2026. 

Multilingual large language models (LLMs) are increasingly deployed in linguistically diverse regions like India, yet most interpretability tools remain tailored to English. Prior work reveals that LLMs often operate in English centric representation spaces, making cross lingual interpretability a pressing concern. We introduce Indic-TunedLens, a novel interpretability framework specifically for Indian languages that learns shared affine transformations. Unlike the standard Logit Lens, which directly decodes intermediate activations, Indic-TunedLens adjusts hidden states for each target language, aligning them with the target output distributions to enable more faithful decoding of model representations. We evaluate our framework on 10 Indian languages using the MMLU benchmark and find that it significantly improves over SOTA interpretability methods, especially for morphologically rich, low resource languages. Our results provide crucial insights into the layer-wise semantic encoding of multilingual transformers. Our model is available at https://huggingface.co/spaces/AnonymousAccountACL/IndicTunedLens. Our code is available at https://github.com/AnonymousAccountACL/IndicTunedLens.

Indic-TunedLens: Interpreting Multilingual Models in Indian Languages

Dialectal speech remains largely underexplored in Automatic Speech Recognition (ASR) research, particularly for Slavic languages. While Ukrainian ASR systems have rapidly improved in recent years with the adoption of Whisper, XLS-R, and Wav2Vec-based models, performance on dialectal variants remains unknown and often significantly degraded. In this work, we present the first dedicated effort to build ASR resources for the Hutsul dialect of Ukrainian. We develop a data preparation and segmentation pipeline, evaluate multiple forced alignment strategies, and benchmark state-of-the-art ASR models under zero-shot and fine-tuned conditions. We evaluate results using WER and CER demonstrating that large multilingual ASR models struggle with dialectal speech, while lightweight fine-tuning produces substantial improvements. All scripts, alignment tools, and training recipes are made publicly available to support future research on Ukrainian dialect speech.

Building ASR Resources for the Hutsul Dialect of Ukrainian

Phonetic transcription is vital for speech processing and linguistic documentation, particularly in languages like Tamil with complex phonology and dialectal variation. Challenges such as consonant gemination, retroflexion, vowel length, and one-to-many grapheme-phoneme mappings are compounded by limited data on Sri Lankan Tamil dialects. We present a dialect-aware, rule-based transcription tool for Tamil that supports Indian and Jaffna Tamil, with extensions underway for other dialects. Using a two-stage pipeline: Tamil script to Latin, then to IPA with context-sensitive rules, the tool handles dialect shifts. A real-time interface enables dialect selection. Evaluated on a 7,830-word corpus, it achieves 94.54\% accuracy for Jaffna Tamil and higher than other tools like eSpeak NG, advancing linguistic preservation and accessible speech technology for Tamil communities.

Bridging Dialectal Variation: A Phonetic Transcription Tool for Tamil

Regional variation was a limiting factor for automatic speech recognition (ASR) before large language models. With the new technology, speech processing becomes more general, which opens the question of how to use data in similar languages such as Croatian and Serbian. In this paper, we analyse model performance in various train-test scenarios with the goal of better understanding the mutual interference of these two languages. Our findings suggest that better performing models are not very sensitive to the regional variation. Training from scratch in one of the languages can give good results on both of them, while fine-tuning large pre-trained multilingual models on smaller data sets does not give the expected results.

Regional Variation in the Performance of ASR Models on Croatian and Serbian

Being modeled as a single-label classification task for a long time, recent work has argued that Arabic Dialect Identification (ADI) should be framed as a multi-label classification task. However, ADI remains constrained by the availability of single-label datasets, with no large-scale multi-label resources available for training. By analyzing models trained on single-label ADI data, we show that the main difficulty in repurposing such datasets for Multi-Label Arabic Dialect Identification (MLADI) lies in the selection of negative samples, as many sentences treated as negative could be acceptable in multiple dialects. To address these issues, we construct a multi-label dataset by generating automatic multi-label annotations using GPT-4o and binary dialect acceptability classifiers, with aggregation guided by the Arabic Level of Dialectness (ALDi). Afterward, we train a BERT-based multi-label classifier using curriculum learning strategies aligned with dialectal complexity and label cardinality. On the MLADI leaderboard, our best-performing LahjatBERT model achieves a macro F1 of 0.69, compared to 0.55 for the strongest previously reported system.

Curriculum Learning and Pseudo-Labeling Improve the Generalization of Multi-Label Arabic Dialect Identification Models

Language identification (LID) is an essential step in building high-quality multilingual datasets from web data. Existing LID tools (such as OpenLID or GlotLID) often struggle to identify closely related languages and to distinguish valid natural language from noise, which contaminates language-specific subsets, especially for low-resource languages. In this work we extend the OpenLID classifier by adding more training data, merging problematic language variant clusters, and introducing a special label for marking noise. We call this extended system OpenLID-v3 and evaluate it against GlotLID on multiple benchmarks. In the evaluation we focus on three groups of closely related languages (Bosnian, Croatian, and Serbian; Romance varieties of Northern Italy and Southern France; and Scandinavian languages) and contribute new evaluation datasets where existing ones are inadequate. Finally, we find that ensemble approaches improve precision but also substantially reduce coverage for low-resource languages.

OpenLID-v3: Improving the Precision of Closely Related Language Identification -- An Experience Report

Despite the success of large language models (LLMs) in a wide range of applications, it has been shown that their performance varies across English dialects. Differences among English dialects are reflected in vocabulary, syntax, and writing style, and can adversely affect model performance. Several studies evaluate the dialect robustness of LLMs, yet research on enhancing their robustness to dialectal variation remains limited.

In this paper, we propose two parameter-efficient frameworks for improving dialectal robustness in LLMs: DialectFusion where we train separate LoRA layers for each dialect and apply different LoRA merging methods, and DialectMoE which is built on top of Mixture of Experts LoRA and introduces multiple LoRA-based experts to the feed-forward layer to internally model the dialectal dependencies. Our comprehensive analysis on five open-source LLMs for sentiment and sarcasm tasks in zero- and few-shot settings shows that our proposed approaches enhance the dialect robustness of LLMs and outperforms instruct and LoRA fine-tuning based approaches.

Improving Dialect Robustness in Large Language Models via LoRA and Mixture-of-Experts

The creation of a robust evaluation methodology is one of the pivotal issues for transfer learning between closely related lects. The current study proposes to resolve this issue by concisely implementing a group of evaluation methods that enable a more systematic qualitative analysis of errata (for instance, string similarity measures to assess lemmatisation more effectively). The paper introduces a robustness score, a metric that aims to assess the stabilityof model performance across different datasets.

The case study is a morphosyntactic tagging of a small historical (beginning of the twentieth century) corpus of Lemko (Slavic clade, Transcarpathian area). It presents a diversity of cross-dependent tasks, made rather complex by the rich Lemko morphology, highly influenced by areal convergence processes. The tagger is a pre-trained Stanza. The study uses modern standard Ukrainian as the source language, as it is the closest to the Lemko high-resource lect.

The analysis reveals that linguistically-aware metrics improve the speed and accuracy of analysis of the errata, especially those caused by the differences between source and target lects. The key data contribution is the open- source dataset of Lemko, obtained during the tagging tasks. Future research directions include a larger-scale test that applies more models to a more extensive material.

Evaluation Framework for Transfer Learning between Closely Related Lects: A Case Study of Lemko

This submission describes a system developed for the AMIYA shared task targeting Syrian Arabic dialect modeling. The approach is based on parameter-efficient fine-tuning using Low-Rank Adaptation (LoRA) applied to a pretrained large language model, combined with prompt-guided inference to encourage fluent and natural dialectal output. The system is designed to prioritize colloquial Syrian Arabic and dialectal fidelity rather than strict factual accuracy, in alignment with the AL-QASIDA evaluation framework. Experimental outputs demonstrate improved fluency and reduced reliance on Modern Standard Arabic compared to the baseline model.

SDNLP at AMIYA 2026: Syrian Arabic Dialect Modeling with LoRA

In this paper, we describe models developed by our team, NUS-IDS, for the Closed data track at the Arabic Modeling In Your Accent (AMIYA) shared task at VarDial 2026. The core idea behind our solution involves data augmentation enabled by a dialect classifier trained on AMIYA data. We effectively combine various translation, summarization, and question answering prompts with AMIYA training data to form dialectal prompts for use with state-of-the-art LLMs. Next, dialect predictions from our classifier on outputs from these LLMs are used to compile preference data for Reinforcement Learning (RL). We report model performance on dialectal Arabic from Egypt, Morocco, Palestine, Saudi Arabia and Syria using FLORES+, a multilingual machine translation dataset. Our experiments illustrate that though our RL models show significant performance gains on dialectness scores, they under perform on translation metrics such as chrF++ compared to base LLMs.

Downloads

Next from EACL 2026

Indic-TunedLens: Interpreting Multilingual Models in Indian Languages

Similar lecture

Arabic Train at NADI 2024 shared task: LLMs’ Ability to Translate Arabic Dialects into Modern Standard Arabic

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from EACL 2026

Indic-TunedLens: Interpreting Multilingual Models in Indian Languages

Similar lecture

Arabic Train at NADI 2024 shared task: LLMs’ Ability to Translate Arabic Dialects into Modern Standard Arabic

Downloads