News is the backbone of democracy; it allows the citizens of a country to be informed about the work of the government. If there is biased news this could affect the citizen’s decision making and hence affect democracy itself. This research aims to create a way to detect biased news articles using a combination of deep learning, word embeddings, and data augmentation techniques. The study utilizes Recurrent Neural Network (RNN), Bidirectional Long Short-Term Memory (BiLSTM), and Contrastive Learning for text classification. Data augmentation techniques have been used to improve the performance of the models. Data augmentation techniques Text Generation, Random Insertion, and Noun Replacement have been used. Word2vec, BERT, RoBERTa, ALBERT, and DeBERTa have been used for creating word embeddings. We are analyzing the impact of different strategies on the accuracy and detection of bias using thorough evaluation and comparison. This study contributes to the advancement of approaches for encouraging objective information transmission and critical thinking when reading news.