LAUSR

Online product configurators, the prevailing toolkits used to realize mass customization, embody an advanced manufacturing strategy that provides customized products with the efficiency of mass production. Essentially, a product configuration system elicits customer needs and maps those needs to product attribute specifications. However, existing configurators require that customers have the necessary domain knowledge to configure their products, which hinders the application of these configurators in current customer-centric product design and manufacturing processes. In this article, we propose a needs-based configurator mechanism that leverages online product-review text from social media. We build a source model that maps product reviews to attribute specifications using a hybrid bidirectional long short-term memory network that incorporates relevant product information at word and character levels. Transfer learning is then deployed to adapt the source model to the target customer needs-specifications mapping. Our experimental results show that the transfer-learning operation significantly improves the configurator performance.