United States

Fake news detection in Malayalam is difficult due to limited data and language challenges. This study compares machine learning, deep learning, and transformer models for classification. The dataset is balanced and divided into training, development and test sets. Machine learning models (SVM, Random Forest, Naive Bayes) used TF-IDF features and deep learning models (LSTM, BiLSTM, CNN) worked with tokenized sequences. We fine-tuned transformer models like IndicBERT, MuRIL, mBERT, and Malayalam-Bert. Among them, the Malayalam-Bert model performed the best and achieved an F1 score of 86%. On the other hand mBERT performed best at spotting fake news. However, the models struggled with mixed-language text and complex writing. Despite these challenges, transformer models turned out to be the most effective for detecting fake news in Malayalam.

NAACL 2025

Celestia@DravidianLangTech 2025: Malayalam-BERT and m-BERT based transformer models for Fake News Detection in Dravidian Languages

dravidian languages

cross-lingual embeddings

fake news detection

shared task

transfer learning

### Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

Welcome to the 2025 meeting of the Nations of the Americas Chapter of the Association for Computational Linguistics! I am proud to help organize the first NAACL conference to carry the new name of our organization, one that emphasizes inclusion for all of the Americas. I am also pleased to welcome you to Albuquerque, New Mexico, a state whose unique blend of cultural influences will make for an excellent backdrop for NAACL 2025, especially with this year’s special theme on NLP in a Multicultural World. 
**[Continue reading...](https://drive.google.com/file/d/1jX-qGhqVSZZCIrAnJaz798pflu5Irrdn/view?usp=sharing)**

*- Colin Cherry, Google, NAACL 2025 General Chair* 

[![](https://assets.underline.io/markdown_image/1/image/b087f8a4dc5816d6e1a6514e59c59ac3.png)](https://drive.google.com/file/d/1T96GzPqObXrMln2BMByCSXTSizjTg69P/view?usp=sharing)

You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Please log in to explore this event.

To access NAACL 2025 event page you are required to register. Please follow [**this link**](https://2025.naacl.org/registration/registration/) to register. Access will depend on your registration type.

Please register!

Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

workshop paper

This paper presents an in-depth study on multi modal hate speech detection in Dravidian lan guages—Tamil, Telugu, and Malayalam—by leveraging both audio and text modalities. De tecting hate speech in these languages is partic ularly challenging due to factors such as code mixing, limited linguistic resources, and di verse cultural contexts. Our approach integrates advanced techniques for audio feature extrac tion and XLM-Roberta for text representation, with feature alignment and fusion to develop a robust multimodal framework. The dataset is carefully categorized into labeled classes: gender-based, political, religious, and personal defamation hate speech, along with a non-hate category. Experimental results indicate that our model achieves a macro F1-score of 0.76 and an accuracy of approximately 85.

Trio Innovators @ DravidianLangTech 2025: Multimodal Hate Speech Detection in Dravidian Languages

Political sentiment analysis has become an essential area of research in Natural Language Processing (NLP), driven by the rapid rise of social media as a key platform for political discourse. This study focuses on sentiment classification in Tamil political tweets, addressing the linguistic and cultural complexities inherent in low-resource languages. To overcome data scarcity challenges, we develop a system that integrates embeddings with advanced Machine Learning techniques, ensuring effective sentiment categorization. Our approach leverages deep learning-based models and transformer architectures to capture nuanced expressions, contributing to improved sentiment classification. This work enhances NLP methodologies for low-resource languages and provides valuable insights into Tamil political discussions, aiding policymakers and researchers in understanding public sentiment more accurately. Notably, our system secured Rank 5 in the NAACL shared task, demonstrating its effectiveness in real-world sentiment classification challenges.

Wictory@DravidianLangTech 2025: Political Sentiment Analysis of Tamil X(Twitter) Comments using LaBSE and SVM

Abusive language directed at women on social media, often characterized by crude slang, offensive terms, and profanity, is not just harmful communication but also acts as a tool for serious and widespread cyber violence. It is imperative that this pressing issue be addressed in order to establish safer online spaces and provide efficient methods for detecting and minimising this kind of abuse. However, the intentional masking of abusive language, especially in regional languages like Tamil and Malayalam, presents significant obstacles, making detection and prevention more difficult. The system created effectively identifies abusive sentences using supervised machine learning techniques based on RoBerta embeddings. The method aims to improve upon the current abusive language detection systems, which are essential for various online platforms, including social media and online gaming services. The proposed method currently ranked 8 in malayalam and 20 in tamil in terms of f1 score.

ANSR@DravidianLangTech 2025: Detection of Abusive Tamil and Malayalam Text Targeting Women on Social Media using RoBERTa and XGBoost

Social media platforms like X (twitter) have gained popularity for political debates and election campaigns in the last decade. This creates the need to moderate and understand the sentiments of the tweets in order to understand the state of digital campaigns. This paper focuses on political sentiment classification of Tamil X (Twitter) comments which proves to be challenging because of the presence of informal expressions, code-switching, and limited annotated datasets. This study focuses on categorizing them into seven classes: substantiated, sarcastic, opinionated, positive, negative, neutral, and none of the above. This paper proposes a solution to Political Multiclass Sentiment Analysis of Tamil X (Twitter) Comments - DravidianLangTech@NAACL 2025 shared task, the solution incorporates IndicBERTv2-MLM-Back-Translation model and TF-IDF vectors into a custom model. Further we explore the use of preprocessing techniques to enrich hashtags and emojis with their context. Our approach achieved Rank 1 with a macro F1 average of 0.38 in the shared task.

Synapse@DravidianLangTech 2025: Multiclass Political Sentiment Analysis in Tamil X (Twitter) Comments: Leveraging Feature Fusion of IndicBERTv2 and Lexical Representations

With the exponential growth of social media usage, the prevalence of abusive language targeting women has become a pressing issue, particularly in low-resource languages (LRLs) like Tamil and Malayalam. This study is part of the shared task at DravidianLangTech@NAACL 2025, which focuses on detecting abusive comments in Tamil social media content. The provided dataset consists of binary-labeled comments (Abusive or Non-Abusive), gathered from YouTube, reflecting explicit abuse, implicit bias, stereotypes, and coded language. We developed and evaluated multiple models for this task, including traditional machine learning algorithms (Logistic Regression, Support Vector Machine, Random Forest Classifier, and Multinomial Naive Bayes), deep learning models (CNN, BiLSTM, and CNN+BiLSTM), and transformer-based architectures (DistilBERT, Multilingual BERT, XLM-RoBERTa), and fine-tuned variants of these models. Our best-performing model, Multilingual BERT, achieved a weighted F1-score of 0.7203, ranking 19 in the competition.

cuetRaptors@DravidianLangTech 2025: Transformer-Based Approaches for Detecting Abusive Tamil Text Targeting Women on Social Media

Zero-shot in-context learning allows large language models (LLMs) to perform tasks using only provided instructions. However, pre-trained LLMs often face calibration issues in zero-shot scenarios, leading to challenges such as hallucinations and off-target translations that compromise output quality, particularly in machine translation (MT). This paper introduces a new method to improve zero-shot MT using fixed prefix pair bootstrapping. By initializing translations with an accurate bilingual prefix pair at the start of both source and target sentences, this approach effectively guides the model to generate precise target-language outputs. Extensive evaluations across four model architectures and multiple translation directions demonstrate significant and consistent improvements, showcasing the potential of this straightforward strategy to enhance zero-shot MT performance.

Enhanced Zero-Shot Machine Translation via Fixed Prefix Pair Bootstrapping

Wenzhou dialect is a Chinese dialect that is significantly distinct from Mandarin, the official language of China, and one of the most complex Chinese dialects. It is nearly incomprehensible to people from regions such as Northern China, creating substantial communication barriers. Therefore, the conversion between Wenzhou dialect and Mandarin is essential to facilitate communication between Wenzhou dialect speakers and people from other regions. However, as a low-resource language, Wenzhou dialect lacks publicly available datasets, and such conversion technologies have not been extensively researched. To address this issue, in this study, we create a parallel dataset containing Wenzhou dialect speech and corresponding Mandarin text and build benchmark models for Wenzhou dialect speech to Mandarin text conversion. In particular, we fine-tune two self-supervised learning-based pre-trained models, TeleSpeech-ASR1.0 and Wav2Vec2-XLS-R, with our training dataset and report their performance on our test dataset as baselines for future research. Our dataset and models are released.

Wenzhou Dialect Speech to Mandarin Text Conversion

Low-resource machine translation (MT) presents a diversity of community needs and application challenges that remain poorly understood. To complement surveys and focus groups, which tend to rely on small samples of respondents, we propose an observational study on actual usage patterns of a specialized MT service for the Tetun language, which is the lingua franca in Timor-Leste. Our analysis of 100,000 translation requests reveals patterns that challenge assumptions based on existing corpora. We find that users, many of them students on mobile devices, typically translate text from a high-resource language into Tetun across diverse domains including science, healthcare, and daily life. This contrasts sharply with available Tetun corpora, which are dominated by news articles covering government and social issues.Our results suggest that MT systems for institutionalized minority languages like Tetun should prioritize accuracy on domains relevant to educational contexts, in the high-resource to low-resource direction. More broadly, this study demonstrates how observational analysis can inform low-resource language technology development, by grounding research in practical community needs.

Low-resource Machine Translation: what for? who for? An observational study on a dedicated Tetun language translation service

Subword tokenization, where text is represented in an intermediate form between full words and characters, is ubiquitous in modern NLP due to its ability to represent any input sentence with a small vocabulary. However for Korean, where there are 11,172 base characters (*syllables*) in its alphabet, it is difficult to have a vocabulary large enough to succinctly encode text while fitting within parameter-budget constraints. This motivates us to explore an alternative representation for Korean which relies on the decompositional nature of Korean syllables: a syllable can be uniquely decomposed into a sequence of two or three subcharacters (*jamo*), of which there are only 68.Using jamo as the basis for subword tokenization (e.g., byte-pair encoding) leads to shorter tokenized sequences with fewer vocabulary parameters, exposes the model to sub-syllable-level morphological information, and increases the amount of augmentation gained from subword regularization. We evaluate jamo-level subword tokenization on several Korean translation tasks and find that jamo-level subword models consistently outperform syllable- and byte-level models in low-resource and restricted-vocabulary settings.

Jamo-Level Subword Tokenization in Low-Resource Korean Machine Translation

We introduce ModeLing, a novel benchmark of Linguistics Olympiad-style puzzles which tests few-shot reasoning in AI systems. Solving these puzzles necessitates inferring aspects of a language's grammatical structure from a small number of examples. Such puzzles provide a natural testbed for language models, as they require compositional generalization and few-shot inductive reasoning. Consisting solely of new puzzles written specifically for this work, ModeLing has no risk of appearing in the training data of existing AI systems: this ameliorates the risk of data leakage, a potential confounder for many prior evaluations of reasoning. Evaluating several large open source language models and GPT on our benchmark, we observe non-negligible accuracy, demonstrating few-shot emergent reasoning ability which cannot merely be attributed to shallow memorization. However, imperfect model performance suggests that ModeLing can be used to measure further progress in linguistic reasoning.

Downloads

Next from NAACL 2025

Trio Innovators @ DravidianLangTech 2025: Multimodal Hate Speech Detection in Dravidian Languages

Similar lecture

Cognitext@DravidianLangTech2025: Fake News Classification in Malayalam Using mBERT and LSTM