Building an Event Knowledge Graph for Telecom Network Operations

The AI R&D Center of China Telecom Beijing Research Institute focuses on developing AI algorithms for network operations, aiming to address growing workload, outdated maintenance models, low data utilization, and limited intelligence in telecom fault management.

To tackle these issues, the center proposes a knowledge graph approach that leverages the strong relational representation of graph databases (Neo4j) to structure heterogeneous data from work orders, case documents, and expert rules, enabling efficient retrieval, reasoning, and faster fault resolution.

The knowledge graph construction process includes four main steps: (1) defining the ontology based on expert collaboration, (2) building the ontology for work order and case data, (3) extracting entities and attributes using the UIE unified information extraction model, and (4) integrating and storing the structured data in Neo4j.

Key extraction tasks involve labeling around 1,000 work orders for fault causes, solutions, alarms, and devices, training the UIE model, and performing entity-level disambiguation. Event extraction further identifies event types, participants, and timestamps to support dynamic recommendation.

Applications of the constructed knowledge graph include the Zhixing Cloud Brain Knowledge Base platform for intelligent search and case management, an AI-powered smart assistant for interactive guidance, and a dynamic recommendation system that suggests fault handling actions during work order processing.

Future directions consider leveraging large language models like ChatGPT to simplify knowledge production, integrate the knowledge graph into LLMs for vertical domain expertise, and further automate network operation tasks, reducing manual effort while maintaining explainable, trustworthy recommendations.