In today’s era, with widespread internet access, social media platforms have become essential for individuals to express their opinions. While sentiment analysis is majorly done in English, there is a notable gap in research for the Hindi language. This paper addresses this problem by proposing three sentiment analysis approaches for Hindi tweets: a Lexicon-based method utilizing Hindi-SentiWordNet, an improved Lexicon-based and a machine learning-based approach using Naïve Bayes Classifier. The research involves creating a Hindi tweet corpus through Twitter API, preprocessing the data, and comparing their performance. Results indicate that the improvised Lexicon-based approach, taking into account positive and negative word counts, yields the best accuracy (65.0%) among all. This paper also addresses the problem of dataset scarcity and provides an efficient data-centric approach rather than basing it on precise polarity scores from resources like H-SWN. This study contributes to enhancing sentiment analysis techniques for the Hindi language in the context of social media communication.