Blockchain, conversation system, master-slave chain, hybrid consensus

Conversation systems are intelligent agents that can help users finish tasks more efficiently via text or spoken interactions, which are among the core technologies in the field of artificial intelligence. The existing conversation systems always need humans or machine learning to maintain the conversation scenarios. To ensure the quality of conversation contents, it is better to use multiple experts group to share and maintain the conversation contents together. In the context of group work, it is important to build fair and trustworthy incentives as rewards. The existing incentive schemes for conversation systems are decided by administrators or managers. Such schemes rely on certain authorities rather than on a consensus for all participants, and the rewards may not be fair or trustworthy. In addition, the contributions are critical for calculating the incentives of all participants, and the conversation contents used to assign the contributions should be protected. In the traditional conversation system, such information is always stored in log files based on a centralized server, which can provide audit trail, but can be easily erased or alterable without trace, and the centralized server also have high privacy risks, providing attackers a single target to hack. Due to these reasons, existing approaches face several issues such as unfair incentive schemes, contributions tampering as well as privacy problems. These inherent fundamental issues in current conversation systems are concerning topics. Blockchain has shown its potential of solving these issues with its key features: autonomous and decentralized processing, smart contractual enforcement of goals, traceable trustworthiness in tamper proof transactions, etc. With the development of the blockchain technology, the value of blockchain lies not only to hold crypto-currencies, but allow in integrating significant panoply information over the same platform in a decentralized and secure way. Although existing projects have opened many doors, but the integration of the conversation system with a blockchain is still in an early prototype stage, and many essential characteristics, such as blockchain interoperability, consensus protocol, and incentive schemes need to be designed and integrated to secure conversation management. Herein, we aim to present a novel blockchain-based decentralized conversation system, that can provide trustworthy and effective conversational services. The key research findings and contributions of this study are: 1) A comprehensive state-of-the-art survey was conducted on the current situations of how the blockchain technology to secure a conversation system and the vision of building a blockchain-based architecture and key technical requirements for building a decentralized conversation system to aid further designing and implementation. 2) Analyzed and identified the requirements of conversation systems and presented a decision model for identifying the best fitting blockchain platform for the conversation systems, which is the foundation of the following designing and development. 3) A novel master‚ÄövÑv¨slave chain model for the conversation system was designed and applied to process multiple conversation interactions concurrently from different domains. 4) A new hybrid consensus algorithm for our master‚ÄövÑv¨slave chain was proposed herein to achieve collaborative sharing and maintenance validation, and an incentive scheme was designed to generate both economic and non-economic rewards for all nodes participating in the proposed consensus process. 5) A decentralized knowledge-fusion scheme with blockchain-enabled smart contracts was implemented based on the proposed blockchain structure, consensus protocol, and incentive scheme. Furthermore, multiple case studies were utilized to evaluate the feasibility and effectiveness of the blockchain-based decentralized conversation system.