Addressing the semantic gap in the consumer-to-manufacturer strategy using dual convolutional neural network

Customer-to-manufacturer (C2M) is an emerging smart manufacturing strategy. In C2M, customers are directly connected with manufacturers for tailor-made product development and manufacturing. Thus, the RD and marketing-driven process in traditional manufacturing evolves into a customer-driven product development process. Product sales using the C2M mode have been one of the highest growth sectors in Chinese ecommerce platforms. However, customers usually lack the necessary domain knowledge. They cannot communicate directly with engineers by indicating the desired technical specifications of the product. A semantic gap exists. This paper presents a dual convolutional neural network-(CNN)-based structure, to automatically address this semantic gap. To mitigate the data sparsity issue in the customer needs domain, we use a massive amount of product review texts which were crawled from ecommerce websites to build a source mapping from reviews to product technical specifications. A small amount of customer needs text was deployed to adapt the source mapping to the target customer needs-product specifications mapping and thus close the semantic gap. Promising experiment results were obtained to show the effectiveness of the method. Copyright © 2021 by IEEE.