Abstract: An embedded information retrieval system is provided including a free-text parser, a knowledge-based query generator, a distributed query dispatcher, and a persistent feedback retrieval manager. When the system is embedded in a text application, the free-text parser takes samples of the text supplied by the user and segments the samples into sentences. The sentences are ranked by their content. The top content-bearing sentences are supplied to the query generator to be converted into queries for the query dispatcher. For each query, the query dispatcher identifies the relevant distributed information sources, submits the query to them, and waits for retrievals. The retrievals are passed to the retrieval manager and saved locally. User feedback is used by the retrieval manager persistently and incrementally to improve retrieval accuracy.