KEC’S CODE CRAFTERS@DravidianLangTech 2026: Abusive Tamil Text Detection Targeting Women on Social Media
Year: 2026
Type: conference-paper
Abstract: Social media is gaining popularity; however, it has become a platform where digital toxicity abounds.Linguistically diverse women find themselves highly vulnerable to cyberbullying.As a result, there is the necessity to develop a solution capable of tackling regionspecific languages.Our paper takes steps in solving this issue by presenting a classification model for "Abusive Tamil Text Detection Targeting Women on Social Media" at the DravidianLangTech-2026 challenge.For developing our model, we used the dataset of 25,948 comments for training and 915 comments for testing.Our main task was the ability to classify content into two classes "Abusive" or "Non-Abusive" for YouTube videos.There were specific challenges associated with using the Tamil language, such as its complicated construction and the fact that the language is often mixed with English within the same sentence.In order to resolve these issues, we designed a pipeline for cleaning these informal scripts.Finally, we ran four traditional ML models: SVM, Logistic Regression, Random Forest, and Multinomial Naive Bayes using TF-IDF as a method of extracting features from texts.Our model achieved the accuracy and F1 scores of 0.86 using Logistic Regression.
Authors: Nivetha, Nethrasri S, Malliga Subramanian
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7166878454