Understanding sentiment expressed in textual data is crucial across various domains, including healthcare, politics, and business. Efficient analysis of large volumes of customer feedback allows companies to assess consumer satisfaction, enhance product quality, and maintain a competitive edge in the market. This increasing need for effective sentiment analysis has led to the emergence of Sentiment Analysis (SA) task. This task aims to classify sentiments in textual data as positive or negative. However, achieving an efficient sentiment classification presents significant challenges, particularly for low-resource languages such as French. In this paper, we address the task of sentiment classification (SC) for the French language by proposing a neural network architecture called CamCnBL (CamemBERT-CNN-Bi-LSTM). This architecture leverages CamemBERT, a transformer-based model, for contextualized embedding generation, Convolutional Neural Networks (CNNs) for local feature extraction, and Bidirectional Long Short-Term Memory (Bi-LSTM) networks to capture contextual relationships. To evaluate the performance of our architecture, we manually annotated a dataset comprising 5,000 French smartphone reviews collected from the Amazon website. Experimental results demonstrate the effectiveness of the proposed architecture, achieving an F-measure of 92.20%. A comparative study further reveals that CamCnBL significantly outperforms other deep learning models, including CNN and Bi-LSTM architectures, as well as other architectures such as CnBR (CNN-Bi-RNN) and CnBG (CNN-Bi-GRU).