United States

Detecting hate speech on social media is increasingly difficult, particularly in low-resource Dravidian languages such as Tamil, Telugu and Malayalam. Traditional approaches primarily rely on text-based classification, often overlooking the multimodal nature of online communication, where speech plays a pivotal role in spreading hate speech. We propose a multimodal hate speech detection model using a late fusion technique that integrates Wav2Vec 2.0 for speech processing and Muril for text analysis. Our model is evaluated on the DravidianLangTech@NAACL 2025 dataset, which contains speech and text data in Telugu, Tamil, and Malayalam scripts. The dataset is categorized into six classes: Non-Hate, Gender Hate, Political Hate, Religious Hate, Religious Defamation, and Personal Defamation. To address class imbalance, we incorporate class weighting and data augmentation techniques. Experimental results demonstrate that the late fusion approach effectively captures patterns of hate speech that may be missed when analyzing a single modality. This highlights the importance of multimodal strategies in enhancing hate speech detection, particularly for low-resource languages.

NAACL 2025

DravLingua@DravidianLangTech 2025: Multimodal Hate Speech Detection in Dravidian Languages using Late Fusion of Muril and Wav2Vec Models

muril_wav2vec

dravidan language

early fusion

late fusion

data augmentation

hate speech detection

### 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

The detection of hate speech in social media platforms is very crucial these days. This is due to its adverse impact on mental health, social harmony, and online safety. This paper presents the overview of the shared task on Multimodal Hate Speech Detection in Dravidian Languages organized as part of DravidianLangTech@NAACL 2025. The task emphasizes detecting hate speech in social media content that combines speech and text. Here, we focus on three low-resource Dravidian languages: Malayalam, Tamil, and Telugu. Participants were required to classify hate speech in three sub-tasks, each corresponding to one of these languages. The dataset was curated by collecting speech and corresponding text from YouTube videos. Various machine learning and deep learning-based models, including transformer-based architectures and multimodal frameworks, were employed by the participants. The submissions were evaluated using the macro F1 score. Experimental results underline the potential of multimodal approaches in advancing hate speech detection for low-resource languages. Team SSNTrio achieved the highest F1 score in Malayalam and Tamil of 0.7511 and 0.7332, respectively. Team lowes scored the best F1 score of 0.3817 in the Telugu sub-task.

Overview of the Shared Task on Multimodal Hate Speech Detection in Dravidian languages: DravidianLangTech@NAACL 2025

The detection of AI-generated product reviews is critical due to the increased use of large language models (LLMs) and their capability to generate convincing sentences. The AIgenerated reviews can affect the consumers and businesses as they influence the trust and decision-making. This paper presents the overview of the shared task on Detecting AI-generated product reviews in Dravidian Languages" organized as part of DravidianLangTech@NAACL 2025. This task involves two subtasks—one in Malayalam and another in Tamil, both of which are binary classifications where a review is to be classified as human-generated or AI-generated. The dataset was curated by collecting comments from YouTube videos. Various machine learning and deep learning-based models ranging from SVM to transformer-based architectures were employed by the participants.

Overview of the Shared Task on Detecting AI generated Product Reviews in Dravidian Languages: DravidianLangTech@NAACL 2025

The rapid expansion of social media has facilitated communication but also enabled the spread of misogynistic memes, reinforcing gender stereotypes and toxic online environments. Detecting such content is challenging due to the multimodal nature of memes, where meaning emerges from the interplay of text and images. The Misogyny Meme Detection shared task at DravidianLangTech@NAACL 2025 focused on Tamil and Malayalam, encouraging the development of multimodal approaches. With 114 teams registered and 23 submitting predictions, participants leveraged various pretrained language models and vision models through fusion techniques. The best models achieved high macro F1 scores (0.83682 for Tamil, 0.87631 for Malayalam), highlighting the effectiveness of multimodal learning. Despite these advances, challenges such as bias in the data set, class imbalance, and cultural variations persist. Future research should refine multimodal detection methods to improve accuracy and adaptability, fostering safer and more inclusive online spaces

Findings of the Shared Task on Misogyny Meme Detection DravidianLangTech@NAACL 2025

Sentiment analysis is an essential task for interpreting subjective opinions and emotions in textual data, with significant implications across commercial and societal applications. This paper provides an overview of the shared task on Sentiment Analysis in Tamil and Tulu, organized as part of DravidianLangTech@NAACL 2025. The task comprises two components: one addressing Tamil and the other focusing on Tulu, both designed as multi-class classification challenges, wherein the sentiment of a given text must be categorized as positive, negative, neutral and unknown. The dataset was diligently organized by aggregating user-generated content from social media platforms such as YouTube and Twitter, ensuring linguistic diversity and real-world applicability. Participants applied a variety of computational approaches, ranging from classical machine learning algorithms such as Traditional Machine Learning Models, Deep Learning Models, Pre-trained Language Models and other Feature Representation Techniques to tackle the challenges posed by linguistic code-mixing, orthographic variations, and resource scarcity in these low resource languages.

Overview of the Shared Task on Sentiment Analysis in Tamil and Tulu

Political multiclass detection is the task of identifying the predefined seven political classes. In this paper, we report an overview of the findings on the "Political Multiclass Sentiment Analysis of Tamil X(Twitter) Comments" shared task conducted at the workshop on DravidianLangTech@NAACL 2025. The participants were provided with annotated Twitter comments, which are split into training, development, and unlabelled test datasets. A total of 139 participants registered for this shared task, and 25 teams finally submitted their results. The performance of the submitted systems was evaluated and ranked in terms of the macro-F1 score.

Overview on Political Multiclass Sentiment Analysis of Tamil X (Twitter) Comments: DravidianLangTech@NAACL 2025

Detecting and mitigating fake news on social media is critical for preventing misinformation, protecting democratic processes, preventing public distress, mitigating hate speech, reducing financial fraud, maintaining information reliability, etc. This paper summarizes the findings of the shared task "Fake News Detection in Dravidian Languages—DravidianLangTech@NAACL 2025." The goal of this task is to detect fake content in social media posts in Malayalam. It consists of two subtasks: the first focuses on binary classification (Fake or Original), while the second categorizes the fake news into five types—False, Half True, Mostly False, Partly False, and Mostly True. In Task 1, 22 teams submitted machine learning techniques like SVM, Naïve Bayes, and SGD, as well as BERT-based architectures. Among these, XLM-RoBERTa had the highest macro F1 score of 89.8%. For Task 2, 11 teams submitted models using LSTM, GRU, XLM-RoBERTa, and SVM. XLM-RoBERTa once again outperformed other models, attaining the highest macro F1 score of 68.2%.

Overview of the Shared Task on Fake News Detection in Dravidian Languages-DravidianLangTech@NAACL 2025

This study introduces a novel multilingual model designed to effectively address the challenges of detecting abusive content in low resource, code-mixed languages, where limited data availability and the interplay of mixed languages, leading to complex linguistic phenomena, create significant hurdles in developing robust machine learning models. By leveraging transfer learning techniques and employing multi-head attention mechanisms, our model demonstrates impressive performance in detecting abusive content in both Tamil and Malayalam datasets. On the Tamil dataset, our team achieved a macro F1 score of 0.7864, while for the Malayalam dataset, a macro F1 score of 0.7058 was attained. These results highlight the effectiveness of our multilingual approach, delivering strong performance in Tamil and competitive results in Malayalam.

Incepto@DravidianLangTech 2025: Detecting Abusive Tamil and Malayalam Text Targeting Women on YouTube

Sentiment analysis on social media platforms plays a crucial role in understanding public opinion and the decision-making process on political matters. As a significant number of individuals express their views on social media, analyzing these opinions is essential for monitoring political trends and assessing voter sentiment. However, sentiment analysis for low-resource languages, such as Tamil, presents considerable challenges due to the limited availability of annotated datasets and linguistic complexities. To address this gap, we utilize a novel dataset encompassing seven sentiment classes, offering a unique opportunity to explore sentiment variations in Tamil political discourse. In this study, we evaluate multiple pre-trained models from the Hugging Face library and experiment with various hyperparameter configurations to optimize model performance. Our findings aim to contribute to the development of more effective sentiment analysis tools tailored for low-resource languages, ultimately empowering Tamil-speaking communities by providing deeper insights into their political sentiments. Our full experimental codebase is publicly available at: ciol-researchlab/NAACL25-Eureka-Sentiment-Analysis-Tamil

Eureka-CIOL@DravidianLangTech 2025: Using Customized BERTs for Sentiment Analysis of Tamil Political Comments

The widespread spread of fake news on social media poses significant challenges, particularly for low-resource languages like Malayalam. The accessibility of social platforms accelerates misinformation, leading to societal polarization and poor decision-making. Detecting fake news in Malayalam is complex due to its linguistic diversity, code-mixing, and dialectal variations, compounded by the lack of large labeled datasets and tailored models. To address these, we developed a fine-tuned transformer-based model for binary and multiclass fake news detection. The binary classifier achieved a macro F1 score of 0.814, while the multiclass model, using multimodal embeddings, achieved a score of 0.1978. Our system ranked 14th and 11th in the shared task competition, highlighting the need for specialized techniques in underrepresented languages. Our full experimental codebase is publicly available at: ciol-researchlab/NAACL25-Akatsuki-Fake-News-Detection.

Akatsuki-CIOL@DravidianLangTech 2025: Ensemble-Based Approach Using Pre-Trained Models for Fake News Detection in Dravidian Languages

Online product reviews influence customer choices and company reputations. However, companies can counter negative reviews by generating fake reviews that portray their products positively. These fake reviews lead to legal disputes and concerns, particularly because AI detection tools are limited in low-resource languages such as Tamil and Malayalam. To address this, we use machine learning and deep learning techniques to identify AI-generated reviews. We utilize Tamil BERT and Malayalam BERT in the embedding layer to extract contextual features. These features are sent to a Feedforward Neural Network (FFN) with softmax to classify reviews as AI-generated or not. The performance of the model is evaluated on the dataset. The results show that the transformer-based embedding achieves a better accuracy of 91.98% on Tamil data and an accuracy of 87.50% on Malayalam data.

Downloads

Next from NAACL 2025

Overview of the Shared Task on Multimodal Hate Speech Detection in Dravidian languages: DravidianLangTech@NAACL 2025

Similar lecture

DLRG@DravidianLangTech 2025: Multimodal Hate Speech Detection in Dravidian Languages