This paper presents a holistic system for semantic annotation, indexing, and retrieval of documents with respect to real-world entities. The system, called KIM, partially implements this concept and is used for evaluation and demonstration. The authors argue that semantic annotation should be based on specific world knowledge rather than general knowledge. A simplistic upper-level ontology is introduced, starting with basic philosophical distinctions and going down to popular entity types, allowing for easy domain-specific extensions. An extensive knowledge base of entity descriptions is maintained based on this ontology. A semantically enhanced information extraction system is presented, providing automatic annotation with references to classes in the ontology and instances in the knowledge base. Based on these annotations, IR-like indexing and retrieval are performed, further extended using the ontology and knowledge about specific entities. The paper discusses the structure and representation of semantic annotations, including necessary knowledge and metadata. It argues for a decoupled representation and management of documents, metadata, and formal knowledge. A light-weight upper-level ontology is advocated for defining entity types, as it allows efficient and scalable management of knowledge. The paper also discusses knowledge representation languages, emphasizing the use of RDF(S) for its widespread acceptance and flexibility. It outlines the principles for metadata encoding and management, including the need for efficient storage and retrieval of annotations. The semantic annotation process is described, focusing on automatic semantic annotation, extraction, indexing, and retrieval. The paper presents the KIM platform, which implements the vision of semantic annotation, indexing, and retrieval. The KIM platform includes a semantic ontology, knowledge base, server, and front-ends for semantic annotation, indexing, and retrieval. The KIM knowledge base is pre-populated with entities of general importance, allowing for effective information extraction and retrieval. The paper also discusses related work, highlighting the importance of semantic annotation in the Semantic Web and the challenges in its implementation. The authors conclude that semantic annotation is essential for the Semantic Web and that further research is needed to develop effective evaluation metrics and techniques for disambiguation of named-entity references.This paper presents a holistic system for semantic annotation, indexing, and retrieval of documents with respect to real-world entities. The system, called KIM, partially implements this concept and is used for evaluation and demonstration. The authors argue that semantic annotation should be based on specific world knowledge rather than general knowledge. A simplistic upper-level ontology is introduced, starting with basic philosophical distinctions and going down to popular entity types, allowing for easy domain-specific extensions. An extensive knowledge base of entity descriptions is maintained based on this ontology. A semantically enhanced information extraction system is presented, providing automatic annotation with references to classes in the ontology and instances in the knowledge base. Based on these annotations, IR-like indexing and retrieval are performed, further extended using the ontology and knowledge about specific entities. The paper discusses the structure and representation of semantic annotations, including necessary knowledge and metadata. It argues for a decoupled representation and management of documents, metadata, and formal knowledge. A light-weight upper-level ontology is advocated for defining entity types, as it allows efficient and scalable management of knowledge. The paper also discusses knowledge representation languages, emphasizing the use of RDF(S) for its widespread acceptance and flexibility. It outlines the principles for metadata encoding and management, including the need for efficient storage and retrieval of annotations. The semantic annotation process is described, focusing on automatic semantic annotation, extraction, indexing, and retrieval. The paper presents the KIM platform, which implements the vision of semantic annotation, indexing, and retrieval. The KIM platform includes a semantic ontology, knowledge base, server, and front-ends for semantic annotation, indexing, and retrieval. The KIM knowledge base is pre-populated with entities of general importance, allowing for effective information extraction and retrieval. The paper also discusses related work, highlighting the importance of semantic annotation in the Semantic Web and the challenges in its implementation. The authors conclude that semantic annotation is essential for the Semantic Web and that further research is needed to develop effective evaluation metrics and techniques for disambiguation of named-entity references.