In the rapidly evolving landscape of global retail, cross-border e-commerce has become a major driver of economic growth, connecting consumers with a vast array of products from around the world. The sheer volume of customer reviews generated on these platforms presents both an opportunity and a challenge for businesses seeking to enhance their product offerings and manufacturing strategies. Effective analysis of these reviews can provide valuable insights into consumer preferences and market trends, ultimately influencing product recommendations and production decisions. This paper proposes an advanced deep learning framework designed to address these challenges by leveraging multi-channel convolutional neural networks (CNNs) and bidirectional gated recurrent units (BiGRUs) for sentiment analysis of product reviews. The CNN component of the model extracts local features at various granularities from the textual data, capturing critical patterns within short sequences of text. However, CNNs alone fall short in modeling the sequential nature of language. To bridge this gap, we integrate BiGRUs, which are adept at learning long-term dependencies and context from sequential data, thus preserving essential information across extended text sequences. To further enhance the model’s performance, we incorporate an attention mechanism following the BiGRU layer. This mechanism dynamically weights the importance of different words in the reviews, allowing the model to focus on the most influential features for classification tasks. By prioritizing key aspects of the text, our approach not only improves the accuracy of product recommendations but also provides more informed guidance for manufacturing decision-making.