Tittle: Cross-border B2B E-commerce Domain Knowledge Graph Construction and Its Application


Knowledge graph (KG) has been widely utilized in information retrieval and reasoning. For a long time, there is a demand for a cross-border B2B e-commerce domain KG that is systematic, renewable and expandable. Such a KG supplies necessary knowledge for ICBU’s transformation into intelligent business decision process, and makes a vital cornerstone of ICBU’s intelligent operations. With the strategic transition of ICBU’s business model, domain KG for sellers’ operation for supporting ICBU sales’ pivoting to consultants, domain KG of cross-border trade full cycle for supporting improving online transaction experience, become two urgent domains of our knowledge construction requirements.


With the sales’ role transitioning to consultants, domain KG of sellers’ operation and its related modules are able to provide highly efficient diagnosis of sellers’ performance and supply accurate solutions accordingly. ICBU Sellers’ most important demand is to obtain new business opportunities and this KG can cover corresponding areas, and provide specific solutions to help them efficiently obtain businesses based on KG’s sophisticated relationships and reasoning modules.  In the strategy of moving offline transactions to online, domain KG of cross-border trade full cycle is able to provide information assistance and risk alert during the complex transaction procedures, which could increase the “perfect order” ratio and decrease dispute ratio.


Existing developed methods mainly focus on constructing general purpose KG by extracting entities and relationship from general semi-structured documents. General-purposed semi-structured documents, such as Wikipedia have been accumulated for a significant amount of time and its quality is also guaranteed. But for domain specific knowledge, for example cross border B2B e-commerce, there is a lack of unified standard and quality assurance, and long-term contribution and maintenance is also not present. Crawling existing semi-structured documents and mining knowledge from them does not suit the construction of domain specific KG. Through motivating and stimulating B2B salesmen/saleswomen, sellers, industry experts and even buyers to contribute knowledge in the formality of crowd sourcing, seems to be the best choice in our scenario. On the basis of analyzing business requirements, we need to clearly define entity and relationship schema, and then develop highly efficient KG editing tool; after obtaining sufficient knowledge, we build knowledge validation, knowledge merging, and knowledge reasoning modules, and integrate the modules into ICBU’s production system to fulfill its functionalities. Constructing and integrating domain KG of sellers’ operation and domain KG of cross-border trade full cycle into our business eco-system, will laydown a solid base for the complete implementation of ICBU’s intelligent decision strategy.