AI Knowledge Graph

### Introduction to Knowledge Graph Construction The construction of knowledge graphs from unstructured documents is a challenging task, especially when dealing with large volumes of heterogeneous data. Large language models can extract entities and relationships from documents, but often do so inconsistently.

### Addressing Inconsistencies with Ontology-Guided Extraction To address these inconsistencies, researchers have developed a production extraction layer that converts a live document stream into a validated knowledge graph aligned to a formal ontology. The system uses a locally hosted Qwen3.5-9B model tuned on the ontology to extract entities and relationships in two passes.

### System Components and Functionality The system consumes document metadata from Kafka and routes content through handlers built for each format, including PDF, spreadsheet, Office, and image content. Its distinguishing component is ontology-guided extraction, which ensures that the extracted entities and relationships are consistent with the formal ontology.

### Conclusion The developed system has the potential to improve the accuracy and consistency of knowledge graph construction from unstructured documents. By utilizing ontology-guided extraction and a production extraction layer, the system can reduce inconsistencies and duplicates, resulting in a more reliable knowledge graph.