Intention model based multi-round dialogue strategies for conversational AI bots

Conversational AI (CoAI) bot, such as customer service bots, navigation bots, chat bots and etc., is a new form of software application. How to accurately identify user requirements and provide appropriate services is one of the biggest challenges that hinder its widespread use. This requires CoAI to get rid of the template and standard knowledge structure, enhance the reasoning ability of service and the knowledge involved. Furthermore, It have the ability to guide users to express complete requirements through multiple rounds of dialogue. To achieve this, firstly, this study purposes a new intention model, and Knowledge Graph of Requirements (KGR) to expand the requirement knowledge scope of conversational AI bots. Secondly, this study purposes a knowledge processing method based on the theory of granular computing. This method could help CoAI bot to determine the next round of inquiry content automatically before the dialogue begins. At last, considering the commonality of the requirement pattern (RP) repository and the personalization of the KGR, this study proposes three optimization methods to improve the dialogue strategy based on two intention models for scenarios with different characteristics. The experimental result shows that, compared with the existing methods, the proposed method can reduce redundancy in dialogues, and integrate the commonality and individuality of user requirements successfully. Meanwhile, this method effectively improves the efficiency and performance of requirement elicitation, and takes the user experience into account as well.