The exponential growth of user-generated content on social networks has necessitated the development of accurate methods for extracting sentiments from text. Sentiment Analysis involves understanding people’s opinions, feelings, and attitudes towards various entities, which is crucial for applications such as improving human–computer interaction, educational tools, and conversational agents. This paper presents deep learning methods that were implemented for sentiment and also has as a special focus the examination of the transparency and interpretability of these deep learning models through the use of LIME. By integrating LIME, we aim to illustrate the decision-making processes of complex deep learning models, providing insights into which features contribute to specific predictions. This approach not only enhances the reliability and fairness of sentiment analysis systems but also builds greater trust in machine learning procedures. The results are quite promising and indicate that the best performance among the deep learning models was achieved by Bi-LSTM which reported a ROC AUC Score of 98.20% and an accuracy of 93.53%.