This research proposes a hybrid deep learning framework for sentiment analysis that integrates pre-trained Global Vectors for Word Representation (GloVe), Convolutional Neural Networks (CNNs), and Bidirectional Long Short-Term Memory (BiLSTM) units. This proposed method is efficient for low-resource systems, thus making it possible for everyone to access. The motivation and novelty of this proposed method lie in the complementary strengths of CNNs in extracting local n-gram features, with the sequential dependency modeling capability of BiLSTM; on the other hand, while GloVe embeddings enrich semantic representation, ultimately showing that a lightweight hybrid model can outperform traditional models using far fewer parameters. The experiments are conducted on the large IMDB benchmark dataset, containing 50,000 balanced movie reviews. Experimental analysis demonstrates that the proposed hybrid framework consistently outperforms individual deep learning models such as RNN, LSTM, CNN, and MLP. Furthermore, comparative analysis with recent studies on the same dataset confirms that the proposed approach achieves higher accuracy and stronger generalization. These results demonstrate the effectiveness of combining pre-trained embeddings with complementary deep learning models, offering a promising direction for advancing sentiment analysis in health monitoring, recommendation systems, and other text-based applications.