Part 3 Building knowledge graphs from text
Transforming unstructured textual data into structured knowledge is an exciting frontier in the development of intelligent systems. This part of the book explores the combination of knowledge graphs (KGs) and large language models (LLMs) in extracting, structuring, and representing knowledge from text, demonstrating how these technologies complement each other to unlock value from unstructured information.
Unstructured text comprises 80% to 90% of enterprise data today. The integration of LLMs has revolutionized this domain, helping us understand and extract meaningful information from content in natural language and reducing the need for human labor. Combining these capabilities with traditional natural language processing (NLP) techniques and KG technologies lets us create systems that can understand context and maintain structured, verifiable knowledge representations.
Chapter 5 demonstrates converting text to KGs and introduces named entity recognition (NER) and relationship extraction using modern LLM-based methods. A case study shows how to extract structured knowledge from historical documents.
Chapter 6 focuses on the workflow from document processing and OCR scanning to graph analytics. We demonstrate a schema design and advanced techniques for data cleaning, entity resolution, and network analysis.