Sentiment analysis on social networking data using NLP and deep learning

One of the most essential methods for tracking public opinion is sentiment analysis of user-generated content on social media, which is crucial for comprehending political discourse. In this study, the ability of Deep Learning models such as CNN, GRU, and LSTM to automatically categorize tweets about the 2020 U.S. Presidential Election is examined. Firstly, tweets were cleaned of any noise, labelled for sentiment, and VADER lexicon was used for normalization. The vocabularies were transformed into word indices and made of equal length so that they could be fed into the neural network models. The accuracy, precision, recall, and F1-score of the models were measured in order to assess their performance. While LSTM and GRU received 68.75% and 74.35% of accuracy, respectively, the CNN model achieved the best accuracy of 85.60% out of all the models, indicating its ability to recognize local textual patterns and key phrases for positive attitudes. The findings indicate that CNNs are very effective in short-text sentiment classification and they outperform recurrent models in managing the nuances of political tweets. The visualization of training curves, confusion matrices, and model comparison metrics also provide further support for the CNN&s;s superiority. The study reveals to what extent Deep Learning (DL) architectures can be successful in political sentiment analysis and suggests that CNN is the most suitable option for dealing with social media conversations.