Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Knowledge Graph Embedding in Intent-Based Networking

Kashif Mehmood, Katina Kralevska, David Palma

TL;DR

This work tackles the complexity of intent-based networking in multi-domain environments by integrating knowledge graphs with a translational Gaussian KG embedding approach. It introduces an Intent Knowledge Graph (IKG) to semantically model intents, services, resources, and KPIs, and adopts Gaussian embeddings (KG2E) to enable predictive completion and verification of intents via a KL-divergence–based scoring function. The authors design a stepwise intent processing pipeline (A–F) that uses the KG embeddings for service prediction, template completion, and triple classification to ensure deployment of only validated intents. Experimental evaluation on a custom dataset demonstrates convergence of the KG2E model and competitive performance in service prediction and intent verification, achieving meaningful accuracy and ranking metrics, and illustrating an end-to-end translation flow from user intent to network action. The resulting pipeline supports dynamic adaptation to network state and resource availability, offering a semantically grounded pathway for reliable service orchestration in 5G/B5G contexts.

Abstract

This paper presents a novel approach to network management by integrating intent-based networking (IBN) with knowledge graphs (KGs), creating a more intuitive and efficient pipeline for service orchestration. By mapping high-level business intents onto network configurations using KGs, the system dynamically adapts to network changes and service demands, ensuring optimal performance and resource allocation. We utilize knowledge graph embedding (KGE) to acquire context information from the network and service providers. The KGE model is trained using a custom KG and Gaussian embedding model and maps intents to services via service prediction and intent validation processes. The proposed intent lifecycle enables intent translation and assurance by only deploying validated intents according to network and resource availability. We evaluate the trained model for its efficiency in service mapping and intent validation tasks using simulated environments and extensive experiments. The service prediction and intent verification accuracy greater than 80 percent is achieved for the trained KGE model on a custom service orchestration intent knowledge graph (IKG) based on TMForum's intent common model.

Knowledge Graph Embedding in Intent-Based Networking

TL;DR

This work tackles the complexity of intent-based networking in multi-domain environments by integrating knowledge graphs with a translational Gaussian KG embedding approach. It introduces an Intent Knowledge Graph (IKG) to semantically model intents, services, resources, and KPIs, and adopts Gaussian embeddings (KG2E) to enable predictive completion and verification of intents via a KL-divergence–based scoring function. The authors design a stepwise intent processing pipeline (A–F) that uses the KG embeddings for service prediction, template completion, and triple classification to ensure deployment of only validated intents. Experimental evaluation on a custom dataset demonstrates convergence of the KG2E model and competitive performance in service prediction and intent verification, achieving meaningful accuracy and ranking metrics, and illustrating an end-to-end translation flow from user intent to network action. The resulting pipeline supports dynamic adaptation to network state and resource availability, offering a semantically grounded pathway for reliable service orchestration in 5G/B5G contexts.

Abstract

This paper presents a novel approach to network management by integrating intent-based networking (IBN) with knowledge graphs (KGs), creating a more intuitive and efficient pipeline for service orchestration. By mapping high-level business intents onto network configurations using KGs, the system dynamically adapts to network changes and service demands, ensuring optimal performance and resource allocation. We utilize knowledge graph embedding (KGE) to acquire context information from the network and service providers. The KGE model is trained using a custom KG and Gaussian embedding model and maps intents to services via service prediction and intent validation processes. The proposed intent lifecycle enables intent translation and assurance by only deploying validated intents according to network and resource availability. We evaluate the trained model for its efficiency in service mapping and intent validation tasks using simulated environments and extensive experiments. The service prediction and intent verification accuracy greater than 80 percent is achieved for the trained KGE model on a custom service orchestration intent knowledge graph (IKG) based on TMForum's intent common model.
Paper Structure (17 sections, 3 equations, 6 figures, 1 table)

This paper contains 17 sections, 3 equations, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Proposed Intent Processing Pipeline with Knowledge Graph Embedding
  3. Integration of Knowledge Graph in Intent Processing
  4. A Simple Knowledge Graph Model
  5. Modeling Intents from IKG
  6. Knowledge Graph Embedding
  7. Training of KG Embedding Model
  8. The Proposed Intent Processing Pipeline
  9. Performance Evaluation
  10. Model Training (KG2E)
  11. Trained KGE Model Performance
  12. Service Predictions
  13. Intent Verification
  14. An Example Intent Processing and Translation Flow
  15. Service Intent Completion
  16. ...and 2 more sections

Figures (6)

  • Figure 1: An Intent Knowledge Graph (IKG) for ibn-kg-self.
  • Figure 2: An example of a .
  • Figure 3: An incomplete service intent template.
  • Figure 4: Proposed intent processing pipeline with .
  • Figure 5: KG2E Model Training Convergence
  • ...and 1 more figures