AIServiceX: A Knowledge Graph-Based Intelligent Question-Answering System for Personal Services

Knowledge graph-based question-answering systems are widely used in e-commerce enterprises. They can reduce the costs of customer services and improve service capabilities. The description of questions is often ambiguous, and the knowledge graph’s update in online personal services always has a high overhead. To address the above issues, by augmenting domain semantics, this paper proposes a knowledge graph-based intelligent question-answering system called as AIServiceX. It employs a gate recurrent unit model to identify entities and assertions, and then gets the most related semantic augmentation contents from existing external domain knowledge via topic comparison. Then, it ranks all the candidates to get optimal answers by discovering several heuristic rules. Periodically, it augments the global knowledge graph with minimized updating costs with an Integer linear programming resolving model. This mechanism can recognize question entities precisely, and map domain knowledge to the KG automatically, which achieves a high answering precision with a low overhead. Experiments with a production e-commerce data show that AIServiceX can improve the precision.