Category-Based Sentiment Analysis of Sindhi News Headlines Using Machine Learning Deep Learning and Transformer Models

The increasing reliance on digital journalism in regional languages has accelerated the demand for automated sentiment analysis tools to evaluate public opinion. This study presents a system for Category-Based Sentiment Analysis of Sindhi News Headlines using a combination of Machine Learning (ML), Deep Learning (DL), and Transformer-based models. The web-based framework enables users to register, upload datasets, and experiment with multiple algorithms for sentiment classification into Positive, Neutral, and Negative categories.

Among ML approaches, Random Forest demonstrated high accuracy and stability, outperforming several traditional classifiers such as Decision Tree and LightGBM. However, the sequential nature of language data proved more suitable for DL models. In particular, the Recurrent Neural Network (RNN) achieved the highest accuracy of 98.87%, establishing it as the most effective model for capturing linguistic dependencies in Sindhi text. The system employs a saved tokenizer with sequence padding, ensuring consistent preprocessing of textual inputs.

Keywords

Sentiment Analysis, Sindhi News Headlines, Random Forest, Recurrent Neural Network (RNN), Machine Learning, Deep Learning, Transformer Models, Natural Language Processing, Text Classification, Public Opinion Mining.