Research on the Construction of Large Model-Driven Intelligent Agents for Waterway Services

With the progressive implementation of the Changjiang River Economic Belt strategy, waterway information has become increasingly multi-source and frequently updated, challenging the capacity of traditional retrieval methods to meet real-time decision-making needs. In response, this study develops an intelligent agent for Changjiang River waterway information services by fine-tuning the QWen model and integrating a waterway knowledge graph to construct a domain-specific large language model. The proposed system performs semantic parsing through the standardisation and categorisation of user queries, accesses relevant data via database connections, knowledge repositories, and intelligent waterway algorithm interfaces, and delivers results in natural language. In comparison with conventional approaches, the agent exhibits markedly improved interaction efficiency, response precision, and domain expertise. Its effectiveness has been verified through practical applications in representative scenarios, providing a promising framework for the intelligent transformation of Changjiang waterway information services and advancing the digitalisation of the shipping industry.