Ontologies can be used to enhance information retrieval and rendition greatly. They also enhance the machine readability and understandability of web documents. Ontologizers and ontology representation languages respectively, can efficiently deduce and describe relationships among information from their metadata in ventures like the Semantic Web. Complete automatic deduction of ontologies from a set of definitive natural language documents, for a specific domain, has not yet been achieved. We describe LOGS, Lightweight universal Ontology Generation and exploitation architectureS, which takes a step toward that direction and the fast, lightweight approach used in it. We also describe the refinement process, heuristics and the new algorithm used in the generation step of LOGS. We illustrate the use of LOGS in 'Eagle', a comprehensive ontology management and information retrieval tool for web documents. Eagle allows the generation and ratification of domain ontologies, and demonstrates their exploitation in intelligent information retrieval and in intelligent interfaces for presenting the information. The tool and LOGS are highly scalable and the ontologies generated can be easily integrated with other tools and domain ontologies, as we use popular ontology languages for representation and processing. Finally, we show how LOGS could be applied in other domains.