Users without knowledge of schemas or structured query languages have difficulties in accessing information stored in databases. Commercial and research efforts have focused on keyword-based searches. Among them, précis queries generate entire multi-relation databases, which are logical subsets of existing ones, instead of individual relations. A logical database subset contains not only items directly related to the query selections but also items implicitly related to them in various ways. Existing approaches to précis query answering assume that a database is pre-annotated with a set of weights, and when a query is issued, an ad-hoc logical subset is constructed on the fly. This approach has several limitations, such as dependence on users for providing appropriate weights and constraints for answering précis queries, and difficulty to capture different query semantics and user preferences. In this paper, we propose a pattern-based approach to logical database subset generation. Patterns of logical subsets corresponding to different queries or user preferences may be recognized and stored in the system. Each time a user poses a question, the system searches in a repository of précis patterns to extract an appropriate one. Then, this is enriched with tuples extracted from the database, in order to produce the logical database subset.