Talk With Your Documents Via Docs2KG
summary
Docs2KG converts diverse enterprise files into a dual-pass pipeline that outputs markdown, OCRed text, tables and figures and assembles a layout + semantic KG so retrieval considers document structure as well as semantics. It targets the claimed 80% of enterprise data in unstructured files, emits Neo4j ingestable JSON, uses OpenAI prompts for entity extraction and implements Microsoft’s GraphRAG for exploratory queries, showing improved one-hop retrieval quality compared with plain RAG.
AI-written summary · about
chapters (6)
- 0:00Introduction to Docs2KG
- 1:56Problem and opportunity with knowledge graphs
- 3:24Handling unstructured data preprocessing
- 7:42Unified pipeline for diverse document formats
- 9:31Building semantic knowledge graphs
- 12:04Evaluation and future work
related talks
01h 00mComplex Data Ingestion with Open Source AI
30mFrom Unstructured Docs to Graph Intelligence: A Framework for Building Graph+Embedding Pipelines
48mIntegrating Docling With OpenSearch for Advanced RAG and Agentic Applications
30mKG Construction Using LLMs
22mStructuring the Unstructured : Empowering the AI Workflows
27mWhy Your RAG is Failing , And How Open Source Can Fix It
32mDocuments Meet LLMs: Tales from the Trenches
01h 30mUnlock Information from Tables, Images and Complex Docs