In the current digital world, users are giving feedback in the form of reviews, blogs, and posts. These opinions are helpful for businesses and organizations to make better decisions. Opinion mining or sentiment analysis is basically used to analyze and classify these opinions in the form of positive, negative, or neutral. This paper proposes an efficient framework that uses natural language processing (NLP) tools and text classification methods to perform opinion mining. The framework includes preprocessing of data including including (tokenization, lemmatization, negation handling, emoji/slang translation), feature extraction, and classification using machine learning algorithms like SVM and Random Forest. The proposed model also integrates the NLP tools like BERT and spaCy for achieving the better accuracy. The datasets from Twitter(short-term) and product reviews(long-term) used to check performance of the proposed system in terms of sentiment classification. This paper also explores the challenges in opinion mining such as sarcasm, spam, and mixed sentiments. This paper shows that by using hybrid NLP and classification techniques can improve the performance and usability of opinion mining systems.