A COMPARATIVE ANALYSIS OF, MACHINE LEARNING AND TRASFORMER MODELS FOR SINDHI NEWS SENTIMENT CLASSIFICATION

Abstract

Sentiment analysis plays a significant role in understanding public opinion and extracting meaningful
insights from textual data, particularly in the field of news and social media analytics. However,
performing sentiment analysis for low-resource languages such as Sindhi remains a challenging task due
to the limited availability of annotated datasets and linguistic resources. This study presents a comparative
analysis of traditional machine learning techniques and modern transformer-based models for Sindhi
news sentiment classification. The research focuses on classifying Sindhi news articles into different
sentiment categories such as positive, negative, and neutral. In the proposed approach, the Sindhi news
dataset is first preprocessed through text cleaning, tokenization, and feature extraction techniques.
Traditional machine learning algorithms such as Naïve Bayes, Support Vector Machines, and Logistic
Regression are applied using feature representation methods like Term Frequency–Inverse Document
Frequency (TF-IDF). In addition, advanced transformer-based models such as BERT and multilingual
transformer architectures are utilized to capture contextual semantic information within the Sindhi
language. The performance of both approaches is evaluated using standard metrics including accuracy,
precision, recall, and F1-score. Experimental results demonstrate that transformer-based models
significantly outperform traditional machine learning methods by providing better contextual
understanding and higher classification accuracy. This study highlights the effectiveness of deep
contextual models in improving sentiment analysis for low-resource languages and contributes to the
development of intelligent natural language processing applications for Sindhi news analytics.