🤖 AI Summary
Existing bibliometric tools exhibit high rigidity, limited flexibility, and steep technical barriers, requiring researchers to possess programming expertise for dynamic scientometric analysis. To address this, we propose the first generative multi-agent AI framework specifically designed for scientometrics, enabling end-to-end analysis via natural language instructions—without coding prerequisites. Our framework innovatively integrates natural language–to–code translation, multimodal full-text retrieval, autonomous agent exploration, and dynamic metric construction. It employs a four-agent collaborative architecture: a Custom Analysis Generator, a Full-Text Retriever, a RAG-powered Research Assistant, and an Automated Report Generator—augmented with sandboxed execution, topic modeling, and embedding-based clustering. Experimental evaluation demonstrates that the system autonomously generates executable scripts, accurately identifies research frontiers, constructs collaboration and citation networks, and produces reproducible, structured scientific reports.
📝 Abstract
Our paper introduces a generative, multiagent AI framework designed to overcome the rigidity, limited flexibility and technical barriers of current bibliometric tools. The objective is to enable researchers to perform fully dynamic, code-based scientometric analysis using natural language NL instructions, eliminating the need for specialized programming skills while expanding analytical depth. Methodologically, the system integrates four coordinated AI agents: a custom analytics generator, a full-paper retriever, including a Retrieval Augmented Generation RAG based researcher assistant and an automated report generator. User queries are translated into executable Python scripts, run within a sandbox ensuring safety, reproducibility and auditability. The framework supports automated data cleaning, construction of co-authorship and citation networks, temporal analyses, topic modeling, embedding based clustering and synthesis of research gaps. Each analytical session produces an exportable, end to end report. The novelty lies in unifying NL to code scientometrics, multimodal full paper retrieval, agentic exploration and dynamic metric creation in a single adaptive environment, capabilities absent in existing platforms: VOSviewer, Bibliometrix, SciMAT. Unlike static GUI based workflows, the proposed framework supports iterative what if analysis, hybrid indicators and user driven pipeline modification. Results demonstrate that the framework generates valid analysis scripts, retrieves and synthesizes full papers, identifies frontier themes and produces reproducible scientometric outputs. It establishes a new paradigm for accessible, interactive and extensible bibliometric knowledge.