Sentiment Analysis of Telugu News Articles Decoding Textual Nuances
V Balaji Viswanadh, Eswar Prasad K, Bharathi Mohan G · 2024
In the field of Natural Language Processing, analyzing sentiment in Telugu presents distinctive challenges owing to the limited availability of annotated datasets. This paper presents a sophisticated method utilizing BERT (Bidirectional Encoder Representations from Transformers) for Telugu sentiment analysis, complemented by Part-of-Speech (POS) and Named Entity Recognition (NER) algorithms. The proposed system follows a multi-step process, leveraging BERT's bidirectional context understanding during pretraining, and subsequently fine-tuning the model with a focus on sentiment analysis. Additionally, the integration of POS and NER algorithms enhances the language model's capabilities by capturing grammatical components and identifying named entities in Telugu text. The model is evaluated on a diverse Telugu dataset, specifically the Telugu Sentiment Analysis Dataset (TSAD), demonstrating its proficiency in predicting sentiment labels. The study addresses the problem of limited annotated data in Telugu sentiment analysis and show-cases significant improvements in accuracy, precision, and recall compared to previous approaches. The proposed system achieves a notable accuracy of 92.07 %, illustrating its effectiveness in capturing the nuances of sentiment expression in Telugu.