Due to the increasing importance of sentiment analysis in analyzing customer feedback, it has been suggested that a framework be developed that combines the various features of deep learning and traditional lexical features. This paper aims to provide a framework that enables companies to perform effective sentiment analysis. The framework is composed of three components: post-processing, feature extraction, and sentiment classification. The first one uses a hybrid approach that combines the traditional features of lexical analysis and contextual features extracted from deep learning models. This method provides a more complete understanding of the sentiments expressed in texts. The second component is sentiment classification, which is performed through a machine learning model that has been trained on hybrid features. The BoW, TF-IDF, Word Embedding perform well in sentiment analysis and outputs a label for each input. The post-processing component of the framework involves analyzing the output of the classification process and applying regression method to improve its accuracy. The results of the study revealed that the hybrid feature extraction technique performed better than the traditional methods when it came to sentiment extraction. The paper also highlighted the advantages of this framework in sentiment analysis by combining the three different techniques