![[2602.12419] Intent-Driven Smart Manufacturing Integrating Knowledge Graphs and Large Language Models](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.12419] Intent-Driven Smart Manufacturing Integrating Knowledge Graphs and Large Language Models
Summary
This article discusses a framework that integrates Large Language Models and Knowledge Graphs to enhance intent-driven interactions in smart manufacturing environments.
Why It Matters
As manufacturing becomes increasingly complex, the ability to translate human intents into machine actions is crucial. This research offers a novel approach that combines advanced AI techniques to improve operational efficiency and adaptability in Manufacturing-as-a-Service ecosystems, which is vital for modern industrial applications.
Key Takeaways
- The framework integrates instruction-tuned LLMs with ontology-aligned KGs for better human-machine interaction.
- Fine-tuning on domain-specific datasets significantly enhances the accuracy of intent translation.
- Achieved 89.33% exact match accuracy and 97.27% overall accuracy in experimental results.
- This approach supports scalable and explainable AI applications in smart manufacturing.
- Aligns with ISA-95 standards, ensuring operational relevance in manufacturing processes.
Computer Science > Artificial Intelligence arXiv:2602.12419 (cs) [Submitted on 12 Feb 2026] Title:Intent-Driven Smart Manufacturing Integrating Knowledge Graphs and Large Language Models Authors:Takoua Jradi, John Violos, Dimitrios Spatharakis, Lydia Mavraidi, Ioannis Dimolitsas, Aris Leivadeas, Symeon Papavassiliou View a PDF of the paper titled Intent-Driven Smart Manufacturing Integrating Knowledge Graphs and Large Language Models, by Takoua Jradi and 6 other authors View PDF HTML (experimental) Abstract:The increasing complexity of smart manufacturing environments demands interfaces that can translate high-level human intents into machine-executable actions. This paper presents a unified framework that integrates instruction-tuned Large Language Models (LLMs) with ontology-aligned Knowledge Graphs (KGs) to enable intent-driven interaction in Manufacturing-as-a-Service (MaaS) ecosystems. We fine-tune Mistral-7B-Instruct-V02 on a domain-specific dataset, enabling the translation of natural language intents into structured JSON requirement models. These models are semantically mapped to a Neo4j-based knowledge graph grounded in the ISA-95 standard, ensuring operational alignment with manufacturing processes, resources, and constraints. Our experimental results demonstrate significant performance gains over zero-shot and 3-shots baselines, achieving 89.33\% exact match accuracy and 97.27\% overall accuracy. This work lays the foundation for scalable, explainable, and adapt...