The spread of erroneous information and fake news on internet media has become a major social issue. This work provides a novel method of identifying false news by combining sentiment-based feature extraction approaches with deep learning models, namely long short-term memory (LSTM) and bidirectional LSTM (Bi-LSTM). Our methodology achieves competitive accuracy in both two-way and six-way classifications on the Politifact dataset by including sentiment analysis into the classification process. Extensive improvements in accuracy and dependability are shown by comparisons with the current literature. By providing insights into the understanding and identification of disinformation in a variety of circumstances, this work advances the development of fake news detection systems.