The potentially conflicting regulations of different nations add risks to cross-border e-commerce (CBEC), and the characteristic of small volume, multi-batch commodity transactions makes CBEC hard to be inspected by human experts. In this paper a knowledge-based CBEC risk assessment system is proposed and special attention is paid to the improvement of system functions including commodity classification and knowledge acquisition. While the former focuses on the accurate recognition of product types in order that right regulations can be applied to the product, the latter accomplishes the transformation from textual regulations into formatted, computer understandable rules with the least mistakes and redundancies. These two techniques collectively improves the usability of the knowledge-based CBEC risk assessment system to a practical level.