Finding a Lattice of Needles in a Haystack: Forming a Query from a Set of Items of Interest

We introduce a new type of query, a lattice query, which is intended to assist a user of a search engine in query formulation. An associated methodology of search is proposed, where instead of submitting an exact query, the user provides a set of text samples which are the basis of generalization. The lattice query system automatically forms a search query from this generalization and verifies the relevancy of search results to the provided set of samples. Lattice queries are also designed to assist in building templates for information extraction tasks: instead of specifying certain keywords or linguistic patterns, a developer can give a list of samples and leave generalization task to the system. Lattice queries are formed from individual sentences and from paragraphs of text as well. An open source contribution of lattice query search component as a part of OpenNLP is described.