Proceedings of the Fourth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages

Bharathi Raja Chakravarthi, Ruba Priyadharshini, Anand Kumar Madasamy, Sajeetha Thavareesan, Elizabeth Sherly, Rajeswari Nadarajan, Manikandan Ravikiran (Editors)

Anthology ID:
St. Julian's, Malta
DravidianLangTech | WS
Association for Computational Linguistics
Bib Export formats:

pdf bib
Proceedings of the Fourth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages
Bharathi Raja Chakravarthi | Ruba Priyadharshini | Anand Kumar Madasamy | Sajeetha Thavareesan | Elizabeth Sherly | Rajeswari Nadarajan | Manikandan Ravikiran

pdf bib
A Few-Shot Multi-Accented Speech Classification for Indian Languages using Transformers and LLM’s Fine-Tuning Approaches
Jairam R | Jyothish G | Premjith B

Accented speech classification plays a vital role in the advancement of high-quality automatic speech recognition (ASR) technology. For certain applications, like multi-accented speech classification, it is not always viable to obtain data with accent variation, especially for resource-poor languages. This is one of the major reasons that contributes to the underperformance of the speech classification systems. Therefore, in order to handle speech variability in Indian language speaker accents, we propose a few-shot learning paradigm in this study. It learns generic feature embeddings using an encoder from a pre-trained whisper model and a classification head for classification. The model is refined using LLM’s fine-tuning techniques, such as LoRA and QLoRA, for the six Indian English accents in the Indic Accent Dataset. The experimental findings show that the accuracy of the model is greatly increased by the few-shot learning paradigm’s effectiveness combined with LLM’s fine-tuning techniques. In optimal settings, the model’s accuracy can reach 94% when the trainable parameters are set to 5%.

pdf bib
Neural Machine Translation for Malayalam Paraphrase Generation
Christeena Varghese | Sergey Koshelev | Ivan P. Yamshchikov

This study explores four methods of generating paraphrases in Malayalam, utilizing resources available for English paraphrasing and pre-trained Neural Machine Translation (NMT) models. We evaluate the resulting paraphrases using both automated metrics, such as BLEU, METEOR, and cosine similarity, as well as human annotation. Our findings suggest that automated evaluation measures may not be fully appropriate for Malayalam, as they do not consistently align with human judgment. This discrepancy underscores the need for more nuanced paraphrase evaluation approaches especially for highly agglutinative languages.

pdf bib
From Dataset to Detection: A Comprehensive Approach to Combating Malayalam Fake News
Devika K | Hariprasath .s.b | Haripriya B | Vigneshwar E | Premjith B | Bharathi Raja Chakravarthi

Identifying fake news hidden as real news is crucial to fight misinformation and ensure reliable information, especially in resource-scarce languages like Malayalam. To recognize the unique challenges of fake news in languages like Malayalam, we present a dataset curated specifically for classifying fake news in Malayalam. This fake news is categorized based on the degree of misinformation, marking the first of its kind in this language. Further, we propose baseline models employing multilingual BERT and diverse machine learning classifiers. Our findings indicate that logistic regression trained on LaBSE features demonstrates promising initial performance with an F1 score of 0.3393. However, addressing the significant data imbalance remains essential for further improvement in model accuracy.

pdf bib
Social Media Fake News Classification Using Machine Learning Algorithm
Girma Bade | Olga Kolesnikova | Grigori Sidorov | José Oropeza

The rise of social media has facilitated easier communication, information sharing, and current affairs updates. However, the prevalence of misleading and deceptive content, commonly referred to as fake news, poses a significant challenge. This paper focuses on the classification of fake news in Malayalam, a Dravidian language, utilizing natural language processing (NLP) techniques. To develop a model, we employed a random forest machine learning method on a dataset provided by a shared task(DravidianLangTech@EACL 2024)1. When evaluated by the separate test dataset, our developed model achieved a 0.71 macro F1 measure.

pdf bib
Exploring the impact of noise in low-resource ASR for Tamil
Vigneshwar Lakshminarayanan | Emily Prud’hommeaux

The use of deep learning algorithms has resulted in significant progress in automatic speech recognition (ASR). Robust high-accuracy ASR models typically require thousands or tens of thousands of hours of speech data, but even the strongest models tend fail under noisy conditions. Unsurprisingly, the impact of noise on accuracy is more drastic in low-resource settings. In this paper, we investigate the impact of noise on ASR in a low-resource setting. We explore novel methods for developing noise-robust ASR models using a a small dataset for Tamil, a widely-spoken but under-resourced Dravidian languages. We add various noises to the audio data to determine the impact of different kinds of noise (e.g., punctuated vs. constant, man-made vs natural) We also explore the relationship between different data augmentation methods are better suited to handling different types of noise. Our results show that all noises, regardless of the type, had an impact on ASR performance, and that upgrading the architecture alone could not mitigate the impact of noise. SpecAugment, the most common data augmentation method, was not as helpful as raw data augmentation, in which noise is explicitly added to the training data. Raw data augmentation enhances ASR performance on both clean data and noise-mixed data.

pdf bib
SetFit: A Robust Approach for Offensive Content Detection in Tamil-English Code-Mixed Conversations Using Sentence Transfer Fine-tuning
Kathiravan Pannerselvam | Saranya Rajiakodi | Sajeetha Thavareesan | Sathiyaraj Thangasamy | Kishore Ponnusamy

Code-mixed languages are increasingly prevalent on social media and online platforms, presenting significant challenges in offensive content detection for natural language processing (NLP) systems. Our study explores how effectively the Sentence Transfer Fine-tuning (Set-Fit) method, combined with logistic regression, detects offensive content in a Tamil-English code-mixed dataset. We compare our model’s performance with five other NLP models: Multilingual BERT (mBERT), LSTM, BERT, IndicBERT, and Language-agnostic BERT Sentence Embeddings (LaBSE). Our model, SetFit, outperforms these models in accuracy, achieving an impressive 89.72%, significantly higher than other models. These results suggest the sentence transformer model’s substantial potential for detecting offensive content in codemixed languages. Our study provides valuable insights into the sentence transformer model’s ability to identify various types of offensive material in Tamil-English online conversations, paving the way for more advanced NLP systems tailored to code-mixed languages.

pdf bib
Findings of the First Shared Task on Offensive Span Identification from Code-Mixed Kannada-English Comments
Manikandan Ravikiran | Ratnavel Rajalakshmi | Bharathi Raja Chakravarthi | Anand Kumar Madasamy | Sajeetha Thavareesan

Effectively managing offensive content is crucial on social media platforms to encourage positive online interactions. However, addressing offensive contents in code-mixed Dravidian languages faces challenges, as current moderation methods focus on flagging entire comments rather than pinpointing specific offensive segments. This limitation stems from a lack of annotated data and accessible systems designed to identify offensive language sections. To address this, our shared task presents a dataset comprising Kannada-English code-mixed social comments, encompassing offensive comments. This paper outlines the dataset, the utilized algorithms, and the results obtained by systems participating in this shared task.

pdf bib
Findings of the Shared Task on Hate and Offensive Language Detection in Telugu Codemixed Text (HOLD-Telugu)@DravidianLangTech 2024
Premjith B | Bharathi Raja Chakravarthi | Prasanna Kumar Kumaresan | Saranya Rajiakodi | Sai Karnati | Sai Mangamuru | Chandu Janakiram

This paper examines the submissions of various participating teams to the task on Hate and Offensive Language Detection in Telugu Codemixed Text (HOLD-Telugu) organized as part of DravidianLangTech 2024. In order to identify the contents containing harmful information in Telugu codemixed social media text, the shared task pushes researchers and academicians to build models. The dataset for the task was created by gathering YouTube comments and annotated manually. A total of 23 teams participated and submitted their results to the shared task. The rank list was created by assessing the submitted results using the macro F1-score.

pdf bib
Findings of the Shared Task on Multimodal Social Media Data Analysis in Dravidian Languages (MSMDA-DL)@DravidianLangTech 2024
Premjith B | Jyothish G | Sowmya V | Bharathi Raja Chakravarthi | K Nandhini | Rajeswari Natarajan | Abirami Murugappan | Bharathi B | Saranya Rajiakodi | Rahul Ponnusamy | Jayanth Mohan | Mekapati Reddy

This paper presents the findings of the shared task on multimodal sentiment analysis, abusive language detection and hate speech detection in Dravidian languages. Through this shared task, researchers worldwide can submit models for three crucial social media data analysis challenges in Dravidian languages: sentiment analysis, abusive language detection, and hate speech detection. The aim is to build models for deriving fine-grained sentiment analysis from multimodal data in Tamil and Malayalam, identifying abusive and hate content from multimodal data in Tamil. Three modalities make up the multimodal data: text, audio, and video. YouTube videos were gathered to create the datasets for the tasks. Thirty-nine teams took part in the competition. However, only two teams, though, turned in their findings. The macro F1-score was used to assess the submissions

pdf bib
Overview of Second Shared Task on Sentiment Analysis in Code-mixed Tamil and Tulu
Lavanya Sambath Kumar | Asha Hegde | Bharathi Raja Chakravarthi | Hosahalli Shashirekha | Rajeswari Natarajan | Sajeetha Thavareesan | Ratnasingam Sakuntharaj | Thenmozhi Durairaj | Prasanna Kumar Kumaresan | Charmathi Rajkumar

Sentiment Analysis (SA) in Dravidian codemixed text is a hot research area right now. In this regard, the “Second Shared Task on SA in Code-mixed Tamil and Tulu” at Dravidian- LangTech (EACL-2024) is organized. Two tasks namely SA in Tamil-English and Tulu- English code-mixed data, make up this shared assignment. In total, 64 teams registered for the shared task, out of which 19 and 17 systems were received for Tamil and Tulu, respectively. The performance of the systems submitted by the participants was evaluated based on the macro F1-score. The best method obtained macro F1-scores of 0.260 and 0.584 for code-mixed Tamil and Tulu texts, respectively.

pdf bib
Overview of the Second Shared Task on Fake News Detection in Dravidian Languages: DravidianLangTech@EACL 2024
Malliga Subramanian | Bharathi Raja Chakravarthi | Kogilavani Shanmugavadivel | Santhiya Pandiyan | Prasanna Kumar Kumaresan | Balasubramanian Palani | Premjith B | Vanaja K | Mithunja S | Devika K | Hariprasath S.b | Haripriya B | Vigneshwar E

The rise of online social media has revolutionized communication, offering users a convenient way to share information and stay updated on current events. However, this surge in connectivity has also led to the proliferation of misinformation, commonly known as fake news. This misleading content, often disguised as legitimate news, poses a significant challenge as it can distort public perception and erode trust in reliable sources. This shared task consists of two subtasks such as task 1 and task 2. Task 1 aims to classify a given social media text into original or fake. The goal of the FakeDetect-Malayalam task2 is to encourage participants to develop effective models capable of accurately detecting and classifying fake news articles in the Malayalam language into different categories like False, Half True, Mostly False, Partly False, and Mostly True. For this shared task, 33 participants submitted their results.

pdf bib
byteSizedLLM@DravidianLangTech 2024: Fake News Detection in Dravidian Languages - Unleashing the Power of Custom Subword Tokenization with Subword2Vec and BiLSTM
Rohith Kodali | Durga Manukonda

This paper focuses on detecting fake news in resource-constrained languages, particularly Malayalam. We present a novel framework combining subword tokenization, Sanskrit-transliterated Subword2vec embeddings, and a powerful Bidirectional Long Short-Term Memory (BiLSTM) architecture. Despite using only monolingual Malayalam data, our model excelled in the FakeDetect-Malayalam challenge, ranking 4th. The innovative subword tokenizer achieves a remarkable 200x compression ratio, highlighting its efficiency in minimizing model size without compromising accuracy. Our work facilitates resource-efficient deployment in diverse linguistic landscapes and sparks discussion on the potential of multilingual data augmentation. This research provides a promising avenue for mitigating linguistic challenges in the NLP-driven battle against deceptive content.

pdf bib
Fida @DravidianLangTech 2024: A Novel Approach to Hate Speech Detection Using Distilbert-base-multilingual-cased
Fida Ullah | Muhammad Zamir | Muhammad Arif | M. Ahmad | E Felipe-Riveron | Alexander Gelbukh

In the contemporary digital landscape, social media has emerged as a prominent means of communication and information dissemination, offering a rapid outreach to a broad audience compared to traditional communication methods. Unfortunately, the escalating prevalence of abusive language and hate speech on these platforms has become a pressing issue. Detecting and addressing such content on the Internet has garnered considerable attention due to the significant impact it has on individuals. The advent of deep learning has facilitated the use of pre-trained deep neural network models for text classification tasks. While these models demonstrate high performance, some exhibit a substantial number of parameters. In the DravidianLangTech@EACL 2024 task, we opted for the Distilbert-base-multilingual-cased model, an enhancement of the BERT model that effectively reduces the number of parameters without compromising performance. This model was selected based on its exceptional results in the task. Our system achieved a commendable Macro F1 score of 0.6369%.

pdf bib
Selam@DravidianLangTech 2024:Identifying Hate Speech and Offensive Language
Selam Abitte Kanta | Grigori Sidorov | Alexander Gelbukh

Social media has transformed into a powerful tool for sharing information while upholding the principle of free expression. However, this open platform has given rise to significant issues like hate speech, cyberbullying, aggression, and offensive language, negatively impacting societal well-being. These problems can even lead to severe consequences such as suicidal thoughts, affecting the mental health of the victims. Our primary goal is to develop an automated system for the rapid detection of offensive content on social media, facilitating timely interventions and moderation. This research employs various machine learning classifiers, utilizing character N-gram TF-IDF features. Additionally, we introduce SVM, RL, and Convolutional Neural Network (CNN) models specifically designed for hate speech detection. SVM utilizes character Ngram TF-IDF features, while CNN employs word embedding features. Through extensive experiments, we achieved optimal results, with a weighted F1-score of 0.77 in identifying hate speech and offensive language.

pdf bib
Tewodros@DravidianLangTech 2024: Hate Speech Recognition in Telugu Codemixed Text
Tewodros Achamaleh | Lemlem Kawo | Ildar Batyrshini | Grigori Sidorov

This study goes into our team’s active participation in the Hate and Offensive Language Detection in Telugu Codemixed Text (HOLDTelugu) shared task, which is an essential component of the DravidianLangTech@EACL 2024 workshop. The ultimate goal of this collaborative work is to push the bounds of hate speech recognition, especially tackling the issues given by codemixed text in Telugu, where English blends smoothly. Our inquiry offers a complete evaluation of the task’s aims, the technique used, and the precise achievements obtained by our team, providing a full insight into our contributions to this crucial linguistic and technical undertaking.

pdf bib
Lidoma@DravidianLangTech 2024: Identifying Hate Speech in Telugu Code-Mixed: A BERT Multilingual
Muhammad Zamir | Moein Tash | Zahra Ahani | Alexander Gelbukh | Grigori Sidorov

Over the past few years, research on hate speech and offensive content identification on social media has been ongoing. Since most people in the world are not native English speakers, unapproved messages are typically sent in code-mixed language. We accomplished collaborative work to identify the language of code-mixed text on social media in order to address the difficulties associated with it in the Telugu language scenario. Specifically, we participated in the shared task on the provided dataset by the Dravidian- LangTech Organizer for the purpose of identifying hate and non-hate content. The assignment is to classify each sentence in the provided text into two predetermined groups: hate or non-hate. We developed a model in Python and selected a BERT multilingual to do the given task. Using a train-development data set, we developed a model, which we then tested on test data sets. An average macro F1 score metric was used to measure the model’s performance. For the task, the model reported an average macro F1 of 0.6151.

pdf bib
Zavira@DravidianLangTech 2024:Telugu hate speech detection using LSTM
Z. Ahani | M. Tash | M. Zamir | I. Gelbukh

Hate speech is communication, often oral or written, that incites, stigmatizes, or incites violence or prejudice against individuals or groups based on characteristics such as race, religion, ethnicity, gender, sexual orientation, or other protected characteristics. This usually involves expressions of hostility, contempt, or prejudice and can have harmful social consequences.Among the broader social landscape, an important problem and challenge facing the medical community is related to the impact of people’s verbal expression. These words have a significant and immediate effect on human behavior and psyche. Repeating such phrases can even lead to depression and social isolation.In an attempt to identify and classify these Telugu text samples in the social media domain, our research LSTM and the findings of this experiment are summarized in this paper, in which out of 27 participants, we obtained 8th place with an F1 score of 0.68.

pdf bib
Tayyab@DravidianLangTech 2024:Detecting Fake News in Malayalam LSTM Approach and Challenges
M. Zamir | M. Tash | Z. Ahani | A. Gelbukh | G. Sidorov

Global communication has been made easier by the emergence of online social media, but it has also made it easier for “fake news,” or information that is misleading or false, to spread. Since this phenomenon presents a significant challenge, reliable detection techniques are required to discern between authentic and fraudulent content. The primary goal of this study is to identify fake news on social media platforms and in Malayalam-language articles by using LSTM (Long Short-Term Memory) model. This research explores this approach in tackling the DravidianLangTech@EACL 2024 tasks. Using LSTM networks to differentiate between real and fake content at the comment or post level, Task 1 focuses on classifying social media text. To precisely classify the authenticity of the content, LSTM models are employed, drawing on a variety of sources such as comments on YouTube. Task 2 is dubbed the FakeDetect-Malayalam challenge, wherein Malayalam-language articles with fake news are identified and categorized using LSTM models. In order to successfully navigate the challenges of identifying false information in regional languages, we use lstm model. This algoritms seek to accurately categorize the multiple classes written in Malayalam. In Task 1, the results are encouraging. LSTM models distinguish between orignal and fake social media content with an impressive macro F1 score of 0.78 when testing. The LSTM model’s macro F1 score of 0.2393 indicates that Task 2 offers a more complex landscape. This emphasizes the persistent difficulties in LSTM-based fake news detection across various linguistic contexts and the difficulty of correctly classifying fake news within the context of the Malayalam language.

pdf bib
IIITDWD_SVC@DravidianLangTech-2024: Breaking Language Barriers; Hate Speech Detection in Telugu-English Code-Mixed Text
Chava Sai | Rangoori Kumar | Sunil Saumya | Shankar Biradar

Social media platforms have become increasingly popular and are utilized for a wide range of purposes, including product promotion, news sharing, accomplishment sharing, and much more. However, it is also employed for defamatory speech, intimidation, and the propagation of untruths about particular groups of people. Further, hateful and offensive posts spread quickly and often have a negative impact on people; it is important to identify and remove them from social media platforms as soon as possible. Over the past few years, research on hate speech detection and offensive content has grown in popularity. One of the many difficulties in identifying hate speech on social media platforms is the use of code-mixed language. The majority of people who use social media typically share their messages in languages with mixed codes, like Telugu–English. To encourage research in this direction, the organizers of DravidianLangTech@EACL-2024 conducted a shared task to identify hateful content in Telugu-English code-mixed text. Our team participated in this shared task, employing three different models: Xlm-Roberta, BERT, and Hate-BERT. In particular, our BERT-based model secured the 14th rank in the competition with a macro F1 score of 0.65.

pdf bib
Beyond Tech@DravidianLangTech2024 : Fake News Detection in Dravidian Languages Using Machine Learning
Kogilavani Shanmugavadivel | Malliga Subramanian | Sanjai R | Mohammed Sameer B | Motheeswaran K

In the digital age, identifying fake news is essential when fake information travels quickly via social media platforms. This project employs machine learning techniques, including Random Forest, Logistic Regression, and Decision Tree, to distinguish between real and fake news. With the rise of news consumption on social media, it becomes essential to authenticate information shared on platforms like YouTube comments. The research emphasizes the need to stop spreading harmful rumors and focuses on authenticating news articles. The proposed model utilizes machine learning and natural language processing, specifically Support Vector Machines, to aggregate and determine the authenticity of news. To address the challenges of detecting fake news in this paper, describe the Machine Learning (ML) models submitted to ‘Fake News Detection in Dravidian Languages” at DravidianLangTech@EACL 2024 shared task. Four different models, namely: Naive Bayes, Support Vector Machine (SVM), Random forest, and Decision tree.

pdf bib
Code_Makers@DravidianLangTech-EACL 2024 : Sentiment Analysis in Code-Mixed Tamil using Machine Learning Techniques
Kogilavani Shanmugavadivel | Sowbharanika J S | Navbila K | Malliga Subramanian

The rising importance of sentiment analysis online community research is addressed in our project, which focuses on the surge of code-mixed writing in multilingual social media. Targeting sentiments in texts combining Tamil and English, our supervised learning approach, particularly the Decision Tree algorithm, proves essential for effective sentiment classification. Notably, Decision Tree(accuracy: 0.99, average F1 score: 0.39), Random Forest exhibit high accuracy (accuracy: 0.99, macro average F1 score : 0.35), SVM (accuracy: 0.78, macro average F1 score : 0.68), Logistic Regression (accuracy: 0.75, macro average F1 score: 0.62), KNN (accuracy: 0.73, macro average F1 score : 0.26) also demonstrate commendable results. These findings showcase the project’s efficacy, offering promise for linguistic research and technological advancements. Securing the 8th rank emphasizes its recognition in the field.

pdf bib
IIITDWD-zk@DravidianLangTech-2024: Leveraging the Power of Language Models for Hate Speech Detection in Telugu-English Code-Mixed Text
Zuhair Shaik | Sai Kartheek Reddy Kasu | Sunil Saumya | Shankar Biradar

Hateful online content is a growing concern, especially for young people. While social media platforms aim to connect us, they can also become breeding grounds for negativity and harmful language. This study tackles this issue by proposing a novel framework called HOLD-Z, specifically designed to detect hate and offensive comments in Telugu-English code-mixed social media content. HOLD-Z leverages a combination of approaches, including three powerful models: LSTM architecture, Zypher, and openchat_3.5. The study highlights the effectiveness of prompt engineering and Quantized Low-Rank Adaptation (QLoRA) in boosting performance. Notably, HOLD-Z secured the 9th place in the prestigious HOLD-Telugu DravidianLangTech@EACL-2024 shared task, showcasing its potential for tackling the complexities of hate and offensive comment classification.

pdf bib
DLRG-DravidianLangTech@EACL2024 : Combating Hate Speech in Telugu Code-mixed Text on Social Media
Ratnavel Rajalakshmi | Saptharishee M | Hareesh S | Gabriel R | Varsini Sr

Detecting hate speech in code-mixed language is vital for a secure online space, curbing harmful content, promoting inclusive communication, and safeguarding users from discrimination. Despite the linguistic complexities of code-mixed languages, this study explores diverse pre-processing methods. It finds that the Transliteration method excels in handling linguistic variations. The research comprehensively investigates machine learning and deep learning approaches, namely Logistic Regression and Bi-directional Gated Recurrent Unit (Bi-GRU) models. These models achieved F1 scores of 0.68 and 0.70, respectively, contributing to ongoing efforts to combat hate speech in code-mixed languages and offering valuable insights for future research in this critical domain.

pdf bib
MIT-KEC-NLP@DravidianLangTech-EACL 2024: Offensive Content Detection in Kannada and Kannada-English Mixed Text Using Deep Learning Techniques
Kogilavani Shanmugavadivel | Sowbarnigaa K S | Mehal Sakthi M S | Subhadevi K | Malliga Subramanian

This study presents a strong methodology for detecting offensive content in multilingual text, with a focus on Kannada and Kannada-English mixed comments. The first step in data preprocessing is to work with a dataset containing Kannada comments, which is backed by Google Translate for Kannada-English translation. Following tokenization and sequence labeling, BIO tags are assigned to indicate the existence and bounds of objectionable spans within the text. On annotated data, a Bidirectional LSTM neural network model is trained and BiLSTM model’s macro F1 score is 61.0 in recognizing objectionable content. Data preparation, model architecture definition, and iterative training with Kannada and Kannada- English text are all part of the training process. In a fresh dataset, the trained model accurately predicts offensive spans, emphasizing comments in the aforementioned languages. Predictions that have been recorded and include offensive span indices are organized into a database.

pdf bib
Transformers@DravidianLangTech-EACL2024: Sentiment Analysis of Code-Mixed Tamil Using RoBERTa
Kriti Singhal | Jatin Bedi

In recent years, there has been a persistent focus on developing systems that can automatically identify the hate speech content circulating on diverse social media platforms. This paper describes the team Transformers’ submission to the Caste/Immigration Hate Speech Detection in Tamil shared task by LT-EDI 2024 workshop at EACL 2024. We used an ensemble approach in the shared task, combining various transformer-based pre-trained models using majority voting. The best macro average F1-score achieved was 0.82. We secured the 1st rank in the Caste/Immigration Hate Speech in Tamil shared task.

pdf bib
Habesha@DravidianLangTech 2024: Detecting Fake News Detection in Dravidian Languages using Deep Learning
Mesay Yigezu | Olga Kolesnikova | Grigori Sidorov | Alexander Gelbukh

This research tackles the issue of fake news by utilizing the RNN-LSTM deep learning method with optimized hyperparameters identified through grid search. The model’s performance in multi-label classification is hindered by unbalanced data, despite its success in binary classification. We achieved a score of 0.82 in the binary classification task, whereas in the multi-class task, the score was 0.32. We suggest incorporating data balancing techniques for researchers who aim to further this task, aiming to improve results in managing a variety of information.

pdf bib
WordWizards@DravidianLangTech 2024:Fake News Detection in Dravidian Languages using Cross-lingual Sentence Embeddings
Akshatha Anbalagan | Priyadharshini T | Niranjana A | Shreedevi Balaji | Durairaj Thenmozhi

The proliferation of fake news in digital media has become a significant societal concern, impacting public opinion, trust, and decision-making. This project focuses on the development of machine learning models for the detection of fake news. Leveraging a dataset containing both genuine and deceptive news articles, the proposed models employ natural language processing techniques, feature extraction and classification algorithms. This paper provides a solution to Fake News Detection in Dravidian Languages - DravidianLangTech 2024. There are two sub tasks: Task 1 - The goal of this task is to classify a given social media text into original or fake. We propose an approach for this with the help of a supervised machine learning model – SVM (Support Vector Machine). The SVM classifier achieved a macro F1 score of 0.78 in test data and a rank 11. The Task 2 is classifying fake news articles in Malayalam language into different categories namely False, Half True, Mostly False, Partly False and Mostly True.We have used Naive Bayes which achieved macro F1-score 0.3517 in test data and a rank 6.

pdf bib
Sandalphon@DravidianLangTech-EACL2024: Hate and Offensive Language Detection in Telugu Code-mixed Text using Transliteration-Augmentation
Nafisa Tabassum | Mosabbir Khan | Shawly Ahsan | Jawad Hossain | Mohammed Moshiul Hoque

Hate and offensive language in online platforms pose significant challenges, necessitating automatic detection methods. Particularly in the case of codemixed text, which is very common in social media, the complexity of this problem increases due to the cultural nuances of different languages. DravidianLangTech-EACL2024 organized a shared task on detecting hate and offensive language for Telugu. To complete this task, this study investigates the effectiveness of transliteration-augmented datasets for Telugu code-mixed text. In this work, we compare the performance of various machine learning (ML), deep learning (DL), and transformer-based models on both original and augmented datasets. Experimental findings demonstrate the superiority of transformer models, particularly Telugu-BERT, achieving the highest f1-score of 0.77 on the augmented dataset, ranking the 1st position in the leaderboard. The study highlights the potential of transliteration-augmented datasets in improving model performance and suggests further exploration of diverse transliteration options to address real-world scenarios.

pdf bib
CUET_Binary_Hackers@DravidianLangTech EACL2024: Fake News Detection in Malayalam Language Leveraging Fine-tuned MuRIL BERT
Salman Farsi | Asrarul Eusha | Ariful Islam | Hasan Mesbaul Ali Taher | Jawad Hossain | Shawly Ahsan | Avishek Das | Mohammed Moshiul Hoque

Due to technological advancements, various methods have emerged for disseminating news to the masses. The pervasive reach of news, however, has given rise to a significant concern: the proliferation of fake news. In response to this challenge, a shared task in Dravidian- LangTech EACL2024 was initiated to detect fake news and classify its types in the Malayalam language. The shared task consisted of two sub-tasks. Task 1 focused on a binary classification problem, determining whether a piece of news is fake or not. Whereas task 2 delved into a multi-class classification problem, categorizing news into five distinct levels. Our approach involved the exploration of various machine learning (RF, SVM, XGBoost, Ensemble), deep learning (BiLSTM, CNN), and transformer-based models (MuRIL, Indic- SBERT, m-BERT, XLM-R, Distil-BERT) by emphasizing parameter tuning to enhance overall model performance. As a result, we introduce a fine-tuned MuRIL model that leverages parameter tuning, achieving notable success with an F1-score of 0.86 in task 1 and 0.5191 in task 2. This successful implementation led to our system securing the 3rd position in task 1 and the 1st position in task 2. The source code will be found in the GitHub repository at this link: DravidianLangTech-EACL-2024-FakeNews.

pdf bib
Punny_Punctuators@DravidianLangTech-EACL2024: Transformer-based Approach for Detection and Classification of Fake News in Malayalam Social Media Text
Nafisa Tabassum | Sumaiya Aodhora | Rowshon Akter | Jawad Hossain | Shawly Ahsan | Mohammed Moshiul Hoque

The alarming rise of fake news on social media poses a significant threat to public discourse and decision-making. While automatic detection of fake news offers a promising solution, research in low-resource languages like Malayalam often falls behind due to limited data and tools. This paper presents the participation of team Punny_Punctuators in the Fake News Detection in Dravidian Languages shared task at DravidianLangTech@EACL 2024, addressing this gap. The shared task focuses on two sub-tasks: 1. classifying social media texts as original or fake, and 2. categorizing fake news into 5 categories. We experimented with various machine learning (ML), deep learning (DL) and transformer-based models as well as processing techniques such as transliteration. Malayalam-BERT achieved the best performance on both sub-tasks, which obtained us 2nd place with a macro f1-score of 0.87 for the subtask-1 and 11th place with a macro f1-score of 0.17 for the subtask-2. Our results highlight the potential of transformer models for low-resource languages in fake news detection and pave the way for further research in this crucial area.

pdf bib
CUET_NLP_GoodFellows@DravidianLangTech EACL2024: A Transformer-Based Approach for Detecting Fake News in Dravidian Languages
Md Osama | Kawsar Ahmed | Hasan Mesbaul Ali Taher | Jawad Hossain | Shawly Ahsan | Mohammed Moshiul Hoque

In this modern era, many people have been using Facebook and Twitter, leading to increased information sharing and communication. However, a considerable amount of information on these platforms is misleading or intentionally crafted to deceive users, which is often termed as fake news. A shared task on fake news detection in Malayalam organized by DravidianLangTech@EACL 2024 allowed us for addressing the challenge of distinguishing between original and fake news content in the Malayalam language. Our approach involves creating an intelligent framework to categorize text as either fake or original. We experimented with various machine learning models, including Logistic Regression, Decision Tree, Random Forest, Multinomial Naive Bayes, SVM, and SGD, and various deep learning models, including CNN, BiLSTM, and BiLSTM + Attention. We also explored Indic-BERT, MuRIL, XLM-R, and m-BERT for transformer-based approaches. Notably, our most successful model, m-BERT, achieved a macro F1 score of 0.85 and ranked 4th in the shared task. This research contributes to combating misinformation on social media news, offering an effective solution to classify content accurately.

pdf bib
CUET_Binary_Hackers@DravidianLangTech EACL2024: Hate and Offensive Language Detection in Telugu Code-Mixed Text Using Sentence Similarity BERT
Salman Farsi | Asrarul Eusha | Jawad Hossain | Shawly Ahsan | Avishek Das | Mohammed Moshiul Hoque

With the continuous evolution of technology and widespread internet access, various social media platforms have gained immense popularity, attracting a vast number of active users globally. However, this surge in online activity has also led to a concerning trend by driving many individuals to resort to posting hateful and offensive comments or posts, publicly targeting groups or individuals. In response to these challenges, we participated in this shared task. Our approach involved proposing a fine-tuning-based pre-trained transformer model to effectively discern whether a given text contains offensive content that propagates hatred. We conducted comprehensive experiments, exploring various machine learning (LR, SVM, and Ensemble), deep learning (CNN, BiLSTM, CNN+BiLSTM), and transformer-based models (Indic-SBERT, m- BERT, MuRIL, Distil-BERT, XLM-R), adhering to a meticulous fine-tuning methodology. Among the models evaluated, our fine-tuned L3Cube-Indic-Sentence-Similarity- BERT or Indic-SBERT model demonstrated superior performance, achieving a macro-average F1-score of 0.7013. This notable result positioned us at the 6th place in the task. The implementation details of the task will be found in the GitHub repository.

pdf bib
TechWhiz@DravidianLangTech 2024: Fake News Detection Using Deep Learning Models
Madhumitha M | Kunguma M | Tejashri J | Jerin Mahibha C

The ever-evolving landscape of online social media has initiated a transformative phase in communication, presenting unprecedented opportunities alongside inherent challenges. The pervasive issue of false information, commonly termed fake news, has emerged as a significant concern within these dynamic platforms. This study delves into the domain of Fake News Detection, with a specific focus on Malayalam. Utilizing advanced transformer models like mBERT, ALBERT, and XMLRoBERTa, our research proficiently classifies social media text into original or fake categories. Notably, our proposed model achieved commendable results, securing a rank of 3 in Task 1 with macro F1 scores of 0.84 using mBERT, 0.56 using ALBERT, and 0.84 using XMLRoBERTa. In Task 2, the XMLRoBERTa model excelled with a rank of 12, attaining a macro F1 score of 0.21, while mBERT and BERT achieved scores of 0.16 and 0.11, respectively. This research aims to develop robust systems capable of discerning authentic from deceptive content, a crucial endeavor in maintaining information reliability on social media platforms amid the rampant spread of misinformation.

pdf bib
CUET_Binary_Hackers@DravidianLangTech-EACL 2024: Sentiment Analysis using Transformer-Based Models in Code-Mixed and Transliterated Tamil and Tulu
Asrarul Eusha | Salman Farsi | Ariful Islam | Jawad Hossain | Shawly Ahsan | Mohammed Moshiul Hoque

Textual Sentiment Analysis (TSA) delves into people’s opinions, intuitions, and emotions regarding any entity. Natural Language Processing (NLP) serves as a technique to extract subjective knowledge, determining whether an idea or comment leans positive, negative, neutral, or a mix thereof toward an entity. In recent years, it has garnered substantial attention from NLP researchers due to the vast availability of online comments and opinions. Despite extensive studies in this domain, sentiment analysis in low-resourced languages such as Tamil and Tulu needs help handling code-mixed and transliterated content. To address these challenges, this work focuses on sentiment analysis of code-mixed and transliterated Tamil and Tulu social media comments. It explored four machine learning (ML) approaches (LR, SVM, XGBoost, Ensemble), four deep learning (DL) methods (BiLSTM and CNN with FastText and Word2Vec), and four transformer-based models (m-BERT, MuRIL, L3Cube-IndicSBERT, and Distilm-BERT) for both languages. For Tamil, L3Cube-IndicSBERT and ensemble approaches outperformed others, while m-BERT demonstrated superior performance among the models for Tulu. The presented models achieved the 3rd and 1st ranks by attaining macro F1-scores of 0.227 and 0.584 in Tamil and Tulu, respectively.

pdf bib
Binary_Beasts@DravidianLangTech-EACL 2024: Multimodal Abusive Language Detection in Tamil based on Integrated Approach of Machine Learning and Deep Learning Techniques
Md. Rahman | Abu Raihan | Tanzim Rahman | Shawly Ahsan | Jawad Hossain | Avishek Das | Mohammed Moshiul Hoque

Detecting abusive language on social media is a challenging task that needs to be solved effectively. This research addresses the formidable challenge of detecting abusive language in Tamil through a comprehensive multimodal approach, incorporating textual, acoustic, and visual inputs. This study utilized ConvLSTM, 3D-CNN, and a hybrid 3D-CNN with BiLSTM to extract video features. Several models, such as BiLSTM, LR, and CNN, are explored for processing audio data, whereas for textual content, MNB, LR, and LSTM methods are explored. To further enhance overall performance, this work introduced a weighted late fusion model amalgamating predictions from all modalities. The fusion model was then applied to make predictions on the test dataset. The ConvLSTM+BiLSTM+MNB model yielded the highest macro F1 score of 71.43%. Our methodology allowed us to achieve 1 st rank for multimodal abusive language detection in the shared task

pdf bib
WordWizards@DravidianLangTech 2024: Sentiment Analysis in Tamil and Tulu using Sentence Embedding
Shreedevi Balaji | Akshatha Anbalagan | Priyadharshini T | Niranjana A | Durairaj Thenmozhi

Sentiment Analysis of Dravidian Languages has begun to garner attention recently as there is more need to analyze emotional responses and subjective opinions present in social media text. As this data is code-mixed and there are not many solutions to code-mixed text out there, we present to you a stellar solution to DravidianLangTech 2024: Sentiment Analysis in Tamil and Tulu task. To understand the sentiment of social media text, we used pre-trained transformer models and feature extraction vectorizers to classify the data with results that placed us 11th in the rankings for the Tamil task and 8th for the Tulu task with a accuracy F1 score of 0.12 and 0.30 which shows the efficiency of our approach.

pdf bib
CUET_DUO@DravidianLangTech EACL2024: Fake News Classification Using Malayalam-BERT
Tanzim Rahman | Abu Raihan | Md. Rahman | Jawad Hossain | Shawly Ahsan | Avishek Das | Mohammed Moshiul Hoque

Identifying between fake and original news in social media demands vigilant procedures. This paper introduces the significant shared task on ‘Fake News Detection in Dravidian Languages - DravidianLangTech@EACL 2024’. With a focus on the Malayalam language, this task is crucial in identifying social media posts as either fake or original news. The participating teams contribute immensely to this task through their varied strategies, employing methods ranging from conventional machine-learning techniques to advanced transformer-based models. Notably, the findings of this work highlight the effectiveness of the Malayalam-BERT model, demonstrating an impressive macro F1 score of 0.88 in distinguishing between fake and original news in Malayalam social media content, achieving a commendable rank of 1st among the participants.

pdf bib
Wit Hub@DravidianLangTech-2024:Multimodal Social Media Data Analysis in Dravidian Languages using Machine Learning Models
Anierudh S | Abhishek R | Ashwin Sundar | Amrit Krishnan | Bharathi B

The main objective of the task is categorised into three subtasks. Subtask-1 Build models to determine the sentiment expressed in multimodal posts (or videos) in Tamil and Malayalam languages, leveraging textual, audio, and visual components. The videos are labelled into five categories: highly positive, positive, neutral, negative and highly negative. Subtask-2 Design machine models that effectively identify and classify abusive language within the multimodal context of social media posts in Tamil. The data are categorized into abusive and non-abusive categories. Subtask-3 Develop advanced models that accurately detect and categorize hate speech and offensive language in multimodal social media posts in Dravidian languages. The data points are categorized into Caste, Offensive, Racist and Sexist classes. In this session, the focus is primarily on Tamil language text data analysis. Various combination of machine learning models have been used to perform each tasks and do oversampling techniques to train models on biased dataset.

pdf bib
CUETSentimentSillies@DravidianLangTech-EACL2024: Transformer-based Approach for Sentiment Analysis in Tamil and Tulu Code-Mixed Texts
Zannatul Tripty | Md. Nafis | Antu Chowdhury | Jawad Hossain | Shawly Ahsan | Avishek Das | Mohammed Moshiul Hoque

Sentiment analysis (SA) on social media reviews has become a challenging research agenda in recent years due to the exponential growth of textual content. Although several effective solutions are available for SA in high-resourced languages, it is considered a critical problem for low-resourced languages. This work introduces an automatic system for analyzing sentiment in Tamil and Tulu code-mixed languages. Several ML (DT, RF, MNB), DL (CNN, BiLSTM, CNN+BiLSTM), and transformer-based models (Indic-BERT, XLM-RoBERTa, m-BERT) are investigated for SA tasks using Tamil and Tulu code-mixed textual data. Experimental outcomes reveal that the transformer-based models XLM-R and m-BERT surpassed others in performance for Tamil and Tulu, respectively. The proposed XLM-R and m-BERT models attained macro F1-scores of 0.258 (Tamil) and 0.468 (Tulu) on test datasets, securing the 2nd and 5th positions, respectively, in the shared task.

pdf bib
Social Media Hate and Offensive Speech Detection Using Machine Learning method
Girma Bade | Olga Kolesnikova | Grigori Sidorov | José Oropeza

Even though the improper use of social media is increasing nowadays, there is also technology that brings solutions. Here, improperness is posting hate and offensive speech that might harm an individual or group. Hate speech refers to an insult toward an individual or group based on their identities. Spreading it on social media platforms is a serious problem for society. The solution, on the other hand, is the availability of natural language processing(NLP) technology that is capable to detect and handle such problems. This paper presents the detection of social media’s hate and offensive speech in the code-mixed Telugu language. For this, the task and golden standard dataset were provided for us by the shared task organizer (DravidianLangTech@ EACL 2024)1. To this end, we have employed the TF-IDF technique for numeric feature extraction and used a random forest algorithm for modeling hate speech detection. Finally, the developed model was evaluated on the test dataset and achieved 0.492 macro-F1.

pdf bib
CUETSentimentSillies@DravidianLangTech EACL2024: Transformer-based Approach for Detecting and Categorizing Fake News in Malayalam Language
Zannatul Tripty | Md. Nafis | Antu Chowdhury | Jawad Hossain | Shawly Ahsan | Mohammed Moshiul Hoque

Fake news misleads people and may lead to real-world miscommunication and injury. Removing misinformation encourages critical thinking, democracy, and the prevention of hatred, fear, and misunderstanding. Identifying and removing fake news and developing a detection system is essential for reliable, accurate, and clear information. Therefore, a shared task was organized to detect fake news in Malayalam. This paper presents a system developed for the shared task of detecting and classifying fake news in Malayalam. The approach involves a combination of machine learning models (LR, DT, RF, MNB), deep learning models (CNN, BiLSTM, CNN+BiLSTM), and transformer-based models (Indic-BERT, XLMR, Malayalam-BERT, m-BERT) for both subtasks. The experimental results demonstrate that transformer-based models, specifically m- BERT and Malayalam-BERT, outperformed others. The m-BERT model achieved superior performance in subtask 1 with macro F1-scores of 0.84, and Malayalam-BERT outperformed the other models in subtask 2 with macro F1- scores of 0.496, securing us the 5th and 2nd positions in subtask 1 and subtask 2, respectively.

pdf bib
MUCS@DravidianLangTech-2024: Role of Learning Approaches in Strengthening Hate-Alert Systems for code-mixed text
Manavi K | Sonali K | Gauthamraj K | Kavya G | Asha Hegde | Hosahalli Shashirekha

Hate and offensive language detection is the task of detecting hate and/or offensive content targetting a person or a group of people. Despite many efforts to detect hate and offensive content on social media platforms, the problem remains unsolved till date due to the ever growing social media users and their creativity to create and spread hate and offensive content. To address the automatic detection of hate and offensive content on social media platforms, this paper describes the learning models submitted by our team - MUCS to “Hate and Offensive Language Detection in Telugu Codemixed Text (HOLD-Telugu): DravidianLangTech@EACL” - a shared task organized at European Chapter of the Association for Computational Linguistics (EACL) 2024 invites the research community to address the challenges of detecting hate and offensive language in Telugu language. In this paper, we - team MUCS, describe the learning models submitted to the above mentioned shared task. Three models: Three models: i) LR model - a Machine Learning (ML) algorithm fed with TF-IDF of n-grams of subword, word and char_wb are in the range (1, 3), (1, 3), and (1, 5), ii) TL- a pretrained BERT models which makes use of Hate-speech-CNERG/bert-base-uncased-hatexplain model and iii) Ensemble model which is the combination of ML classifieres( MNB, LR, GNB) trained CountVectorizer with word and char ngrams of range (1, 3) and (1, 5) respectively. Proposed LR model trained with TF-IDF of subword, word and char n-grams outperformed the other models with macro F1 scores of 0.6501 securing 15th rankin the shared task for Telugu text.

pdf bib
MUCS@DravidianLangTech-2024: A Grid Search Approach to Explore Sentiment Analysis in Code-mixed Tamil and Tulu
Prathvi B | Manavi K | Subrahmanyapoojary K | Asha Hegde | Kavya G | Hosahalli Shashirekha

Sentiment Analysis (SA) is a field of computational study that analyzes and understands people’s opinions, attitudes, and emotions toward any entity. A review of an entity can be written about an individual, an event, a topic, a product, etc., and such reviews are abundant on social media platforms. The increasing number of social media users and the growing amount of user-generated code-mixed content such as reviews, comments, posts etc., on social media have resulted in a rising demand for efficient tools capable of effectively analyzing such content to detect the sentiments. In spite of this, SA of social media text is challenging because the code-mixed text is complex. To address SA in code-mixed Tamil and Tulu text, this paper describes the Machine Learning (ML) models submitted by our team - MUCS to “Sentiment Analysis in Tamil and Tulu - Dravidian- LangTech” - a shared task organized at European Chapter of the Association for Computational Linguistics (EACL) 2024. Linear Support Vector classifier (LinearSVC) and ensemble of 5 ML classifiers (k Nearest Neighbour (kNN), Stochastic Gradient Descent (SGD), Logistic Regression (LR), LinearSVC, and Random Forest Classifier (RFC)) with hard voting trained using concatenated features obtained from word and character n-ngrams vectoized from Term Frequency-Inverse Document Frequency (TF-IDF) vectorizer and CountVectorizer. Further, Gridsearch algorithm is employed to obtain optimal hyperparameter values.The proposed ensemble model obtained macro F1 scores of 0.260 and 0.550 for Tamil and Tulu languages respectively.

pdf bib
InnovationEngineers@DravidianLangTech-EACL 2024: Sentimental Analysis of YouTube Comments in Tamil by using Machine Learning
Kogilavani Shanmugavadivel | Malliga Subramanian | Palanimurugan V | Pavul chinnappan D

There is opportunity for machine learning and natural language processing research because of the growing volume of textual data. Although there has been little research done on trend extraction from YouTube comments, sentiment analysis is an intriguing issue because of the poor consistency and quality of the material found there. The purpose of this work is to use machine learning techniques and algorithms to do sentiment analysis on YouTube comments pertaining to popular themes. The findings demonstrate that sentiment analysis is capable of giving a clear picture of how actual events affect public opinion. This study aims to make it easier for academics to find high-quality sentiment analysis research publications. Data normalisation methods are used to clean an annotated corpus of 1500 citation sentences for the study. .For classification, a system utilising one machine learning algorithm—K-Nearest Neighbour (KNN), Na ̈ıve Bayes, SVC (Support Vector Machine), and RandomForest—is built. Metrics like the f1-score and correctness score are used to assess the correctness of the system.

pdf bib
KEC_HAWKS@DravidianLangTech 2024 : Detecting Malayalam Fake News using Machine Learning Models
Malliga Subramanian | Jayanthjr J R | Muthu Karuppan P | Keerthibala T | Kogilavani Shanmugavadivel

The proliferation of fake news in the Malayalam language across digital platforms has emerged as a pressing issue. By employing Recurrent Neural Networks (RNNs), a type of machine learning model, we aim to distinguish between Original and Fake News in Malayalam and achieved 9th rank in Task 1.RNNs are chosen for their ability to understand the sequence of words in a sentence, which is important in languages like Malayalam. Our main goal is to develop better models that can spot fake news effectively. We analyze various features to understand what contributes most to this accuracy. By doing so, we hope to provide a reliable method for identifying and combating fake news in the Malayalam language.