BPsim Decision System and Twin Intelligent Language Processing: Developing Domain-Specific Expert Systems

As an essential tool for human-computer interaction, question-answering systems provide efficient support for humans in addressing complex work demands. From the rule-based simple question-answering systems of the 1950s to the current large language models incorporating the Transformer architecture, question-answering systems have made significant progress in understanding natural language and generating accurate responses. However, existing large language models still face challenges such as unstable outputs, complex fine-tuning, and high computational demands when applied in specific domains. This paper proposes a novel question-answering system architecture that integrates the TWIN intelligent language recognition system and the BPsim.DSS expert decision-making system, aiming to enhance the natural language understanding capabilities of question-answering systems and improve their application effectiveness in providing high-confidence decision results in professional fields. The system's efficiency and practicality in the airline service domain are validated through partial airline data. This system not only responds quickly to user needs but also provides precise and flexible solutions through the integration of expert systems and text analysis, offering new ideas and methods for human-computer collaboration and complex task processing.