Abstract: Data pertaining to a subject matter domain, a set of text strings forming a set of seeds, a description of a linguistic structure present in a language of the domain-related data, and a statistical model applicable to the domain-related data are received. A set of portions of the domain-related data is extracted, a portion in the set of portions forming a nugget. A nugget matches the statistical model according to a criterion, and conforms to the linguistic structure within a threshold degree. The nugget is scored according to a subset of a set of features found in the nuggets. A subset of nuggets is selected. A score of each nugget included in the subset of nuggets exceeds a score threshold. The subset of nuggets is combined to form a pseudo-document. The pseudo-document is submitted to an application for answering a question related to the domain.

Abstract: Data pertaining to a subject matter domain, a set of text strings forming a set of seeds, a description of a linguistic structure present in a language of the domain-related data, and a statistical model applicable to the domain-related data are received. A set of portions of the domain-related data is extracted, a portion in the set of portions forming a nugget. A nugget matches the statistical model according to a criterion, and conforms to the linguistic structure within a threshold degree. The nugget is scored according to a subset of a set of features found in the nuggets. A subset of nuggets is selected. A score of each nugget included in the subset of nuggets exceeds a score threshold. The subset of nuggets is combined to form a pseudo-document. The pseudo-document is submitted to an application for answering a question related to the domain.