There has been much work in the past that combines the fields of Databases and Natural Language Processing. Almost all efforts, however, have gone in one direction: given unstructured, natural-language elements (requests, text excerpts, etc.), one creates structured, database elements (queries, records, etc.). The other direction has been mostly ignored, although it is very rich in application opportunities as well as in research problems. This short paper outlines several aspects of this other direction, identifying some relevant technical challenges as well as corresponding real applications.