Document templates could improve the speed, cost, and quality of documentation if appropriate templates could be located without undue selection burden. Semantic retrieval (IR) can greatly improve the precision of finding relevant document templates. The present invention discloses a process which implements a semantic information retrieval system to locate template documents, using as search vectors the semantic content of sentences from a new partially completed document. The system enables the author to quickly and easily transfer sentences from template documents into the new document. The system provides options for the author to match against specialized template collections and subsets of template documents. A significant advantage of the present invention over other template based methods is ability to retrieve a relevant template document when there are many thousands of exemplar documents without having to construct a formal query.