Research on the Evolution Analysis Framework of Keywords for Intelligent Agents Oriented to Monitoring the Frontiers of Disciplines

Abstract

Promptly identify and study new-generation disciplines to provide support for technological strategy and resource allocation. Research has shown that large language models can be used to study literature on intelligent agents to some extent. The problems of keyword evolution analysis, multi-source data integration and automation have not been solved by the current methods. A framework for the evolution analysis of keywords is put forward in this paper to track the leading edge of disciplines based on the evolution analysis of keywords and the technological progress of intelligent agents. The framework is based on intelligent agents, collects data from multiple sources, and sets up four modules for extraction and standardisation, topic modelling and clustering, evolution identification and signal detection, and thus achieves full-process analysis of the data from collection to result output. A multi-granularity tracking and multi-dimensional feature fusion method are used by the framework to improve the timeliness, accuracy and interpretability of monitoring. Intelligent agents can autonomously complete complex tasks and perform better than a single model; according to CAICT's 2025 industry report, fully packaged general intelligent agents outperform some top-level large models, and the core is to build a closed-loop system via dynamic planning engines and tool calls. Intelligent agents can be introduced to correct the shortcomings of traditional measurement methods in dynamic characterization and intelligence level, and a new type of monitoring for all areas can be provided.