US11238864B2 - Selectively Generating Expanded Responses That Guide Continuance of a Human-to-Computer Dialog