Austria

In this paper we introduce the first effort to adapt large language models (LLMs) to the Ukrainian dialect (in our case Hutsul), a low-resource and morphologically complex dialect spoken in the Carpathian Highlands. We created a parallel corpus of 9852 dialect-to-standard Ukrainian sentence pairs and a dictionary of 7320 dialectal word mappings. We also addressed data shortage by proposing an advanced Retrieval-Augmented Generation (RAG) pipeline to generate synthetic parallel translation pairs, expanding the corpus with 52142 examples. We have fine-tuned multiple open-source LLMs using LoRA and evaluated them on a standard-to-dialect translation task, also comparing with few-shot GPT-4o translation. In the absence of human annotators, we adopt a multi-metric evaluation strategy combining BLEU, chrF++, TER, and LLM-based judgment (GPT-4o). The results show that even small(7B) finetuned models outperform zero-shot baselines such as GPT-4o across both automatic and LLM-evaluated metrics. All data, models, and code are publicly released at: https://github.com/woters/vuyko-hutsul.

ACL 2025

Vuyko Mistral: Adapting LLMs for Low-Resource Dialectal Translation

#### *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

EACL 2026

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. 

workshop paper

### Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Message from the General Chair: 
*It is my great pleasure and honor to welcome you to the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), held in beautiful Vienna, Austria, from July 27 to August 1, 2025. ACL2025continues our field’s tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.* 
[Read more](https://drive.google.com/file/d/1GI_hvOpjswAuYdUTromfeDiPpCcqidwg/view?usp=sharing)

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 ACL 2025". Please be sure to check your spam email folder if you do not see an email confirmation right away.

To access this event page, you need to log in with the **email address you registered with**. 

Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Text preprocessing is a fundamental component of high-quality speech synthesis. This work presents a novel rule-based phonemizer combined with a sentence-level lexical stress prediction model to improve phonetic accuracy and prosody prediction in the text-to-speech pipelines. We also introduce a new benchmark dataset with annotated stress patterns designed for evaluating lexical stress prediction systems at the sentence level.

Experimental results demonstrate that the proposed phonemizer achieves a 1.23% word error rate on a manually constructed pronunciation dataset, while the lexical stress prediction pipeline shows results close to dictionary-based methods, outperforming existing neural network solutions.

Context-Aware Lexical Stress Prediction and Phonemization for Ukrainian TTS Systems

Synonym-based adversarial tests reveal fragile word patterns that accuracy metrics overlook, while virtually no such diagnostics exist for Ukrainian, a morphologically rich and low‑resource language. We present the first systematic robustness evaluation under synonym substitution in Ukrainian. Adapting \textsc{TextFooler} and \textsc{BERT‑Attack} to Ukrainian, we (i) adjust a 15000‑entry synonym dictionary to match proper word forms; (ii) integrate similarity filters; (iii) adapt masked‑LM search so it generates only valid inflected words. Across three text classification datasets (reviews, news headlines, social‑media manipulation) and three transformer models (Ukr‑RoBERTa, XLM‑RoBERTa, SBERT), single‑word swaps reduce accuracy by up to 12.6, while multi‑step attacks degrade performance by as much as 40.27 with around 112 model queries. A few‑shot transfer test shows GPT‑4o, a state‑of‑the‑art multilingual LLM, still suffers 6.9–15.0 drops on the same adversarial samples. Our results underscore the need for sense‑aware, morphology‑constrained synonym resources and provide a reproducible benchmark for future robustness research in Ukrainian NLP.

Precision vs. Perturbation: Robustness Analysis of Synonym Attacks in Ukrainian NLP

This paper introduces UAlign, the comprehensive benchmark for evaluating the alignment of Large Language Models (LLMs) in the Ukrainian language. The benchmark consists of two complementary components: a moral judgment dataset with 3,682 scenarios of varying ethical complexities and a dataset with 1,700 ethical situations presenting clear normative distinctions. Each element provides parallel English-Ukrainian text pairs, enabling cross-lingual comparison. Unlike existing resources predominantly developed for high-resource languages, our benchmark addresses the critical need for evaluation resources in Ukrainian. The development process involved machine translation and linguistic validation using Ukrainian language models for grammatical error correction. Our cross-lingual evaluation of six LLMs confirmed the existence of a performance gap between alignment in Ukrainian and English while simultaneously providing valuable insights regarding the overall alignment capabilities of these models. The benchmark has been made publicly available to facilitate further research initiatives and enhance commercial applications.

Warning: The datasets introduced in this paper contain sensitive materials related to ethical and moral scenarios that may include offensive, harmful, illegal, or controversial content.

UAlign: LLM Alignment Benchmark for the Ukrainian Language

Large Language Models (LLMs) have demonstrated remarkable performance across various domains, but they often inherit biases present in the data they are trained on, leading to unfair or unreliable outcomes—particularly in sensitive areas such as hiring, medical decision-making, and education. This paper evaluates gender bias in LLMs within the Ukrainian language context, where the gendered nature of the language and the use of feminitives introduce additional complexity to bias analysis. We propose a benchmark for measuring bias in Ukrainian and assess several debiasing methods, including prompt debiasing, embedding debiasing, and fine-tuning, to evaluate their effectiveness. Our results suggest that embedding debiasing alone is insufficient for a morphologically rich language like Ukrainian, whereas fine-tuning proves more effective in mitigating bias for domain-specific tasks.

GBEM-UA: Gender Bias Evaluation and Mitigation for Ukrainian Large Language Models


This paper presents a pipeline for generating gender-balanced datasets through sentence-level gender swapping, addressing the gender-imbalance issue in Ukrainian texts. We select sentences with gender-marked entities, focusing on job titles, generate their inverted alternatives using LLMs and human-in-the-loop, and fine-tune Aya-101 on the resulting dataset for the task of gender swapping. Additionally, we train a Named Entity Recognition (NER) model on gender-balanced data, demonstrating its ability to better recognize gendered entities. The findings unveil the potential of gender-balanced datasets to enhance model robustness and support more fair language processing. Finally, we make a gender-swapped version of NER-UK~2.0 and the fine-tuned Aya-101 model available for download and further research.

Gender Swapping as a Data Augmentation Technique: Developing Gender-Balanced Datasets for Ukrainian Language Processing

This paper presents the results of the UNLP 2025 Shared Task on Detecting Social Media Manipulation. The task included two tracks: Technique Classification and Span Identification. 
The benchmark dataset contains 9,557 posts from Ukrainian Telegram channels manually annotated by media experts. A total of 51 teams registered, 22 teams submitted systems, and 595 runs were evaluated on a hidden test set via Kaggle. Performance was measured with macro F1 for classification and token‑level F1 for identification. 
The shared task provides the first publicly available benchmark for manipulation detection in Ukrainian social media and highlights promising directions for low‑resource propaganda research. The Kaggle leaderboard is left open for further submissions.

The UNLP 2025 Shared Task on Detecting Social Media Manipulation

The Russia-Ukraine war has transformed social media into a critical battleground for information warfare, making the detection of manipulation techniques in online content an urgent security concern. This work presents our system developed for the UNLP 2025 Shared Tasks, which addresses both manipulation technique classification and span identification in Ukrainian Telegram posts. In this paper, we have explored several machine learning approaches (LR, SVC, GB, NB) , deep learning architectures (CNN, LSTM, BiLSTM, GRU hybrid) and state-of-the-art multilingual transformers (mDeBERTa, InfoXLM, mBERT, XLM-RoBERTa). Our experiments showed that fine-tuning transformer models for the specific tasks significantly improved their performance, with XLM-RoBERTa large delivering the best results by securing 3rd place in technique classification task with a Macro F1 score of 0.4551 and 2nd place in span identification task with a span F1 score of 0.6045. These results demonstrate that large pre-trained multilingual models effectively detect subtle manipulation tactics in Slavic languages, advancing the development of tools to combat online manipulation in political contexts.

Detecting Manipulation in Ukrainian Telegram: A Transformer-Based Approach to Technique Classification and Span Identification

This paper presents one of the top-performing solutions to the UNLP 2025 Shared Task on Detecting Manipulation in Social Media. The task focuses on detecting and classifying rhetorical and stylistic manipulation techniques used to influence Ukrainian Telegram users. For the classification subtask, we fine-tuned the Gemma 2 language model with LoRA adapters and applied a second-level classifier leveraging meta-features and threshold optimization. For span detection, we employed an XLM-RoBERTa model trained for multi-target, including token binary classification. Our approach achieved 2nd place in classification and 3rd place in span detection.

Hidden Persuasion: Detecting Manipulative Narratives on Social Media During the 2022 Russian Invasion of Ukraine

Detecting manipulation techniques in online text is vital for combating misinformation, a task complicated by generative AI. This paper compares machine learning approaches for multi-label classification of 10 techniques in Ukrainian Telegram content (UNLP 2025 Shared Task 1). Our evaluation included TF-IDF, fine-tuned XLM-RoBERTa-Large, PEFT-LLM (Gemma, Mistral) and a RAG approach (E5 + Mistral Nemo). The fine-tuned XLM-RoBERTa-Large model, which incorporates weighted loss to address class imbalance, yielded the highest Macro F1 score (0.4346). This result surpassed the performance of TF-IDF (Macro F1 0.32-0.36), the PEFT-LLM (0.28-0.33) and RAG (0.309). Synthetic data slightly helped TF-IDF but reduced transformer model performance. The results demonstrate the strong performance of standard transformers like XLM-R when appropriately configured for this classification task.

Comparing Methods for Multi-Label Classification of Manipulation Techniques in Ukrainian Telegram Content

In this paper, we present our solutions for the two UNLP 2025 shared tasks: manipulation span detection and manipulation technique classification in Ukraine-related media content sourced from Telegram channels. 

We experimented with fine-tuning large language models (LLMs) with up to 12 billion parameters, including both encoder- and decoder-based architectures. Our experiments identified Gemma 3 12b with a custom classification head as the best-performing model for both tasks. 

To address the limited size of the original training dataset, we generated 50k synthetic samples and marked up an additional 400k media entries containing manipulative content.

Downloads

Next from ACL 2025

Context-Aware Lexical Stress Prediction and Phonemization for Ukrainian TTS Systems

Similar lecture

Building ASR Resources for the Hutsul Dialect of Ukrainian