Abstract: As a major integrative academic output of higher education, dissertations hold significant scholarly value. This study aims to effectively organize and deeply mine the knowledge contained within dissertations, constructing a knowledge service framework for dissertations to fully unleash their intrinsic value. It follows a technical pathway of systematic analysis, intelligent analysis, and service-oriented design. Firstly, it comprehensively examines the structural features and types of knowledge units in dissertations, delineating the boundaries of content elements that can be extracted, organized, and served. Secondly, it proposes a large language model (LLM)-driven framework for dissertation knowledge services, exploring intelligent analysis methods based on LLMs and designing a scenario-driven functional system for dissertation knowledge services, covering retrieval, analysis, recommendation, and visualization, among others. Finally, by developing a prototype system, the study validates the effectiveness and practical value of the proposed intelligent analysis approach and knowledge service framework, providing a reference for the deep development and utilization of dissertation resources.

Key words: Dissertation, Knowledge unit, Intelligent analysis, Knowledge service, Knowledge mining