Patent 9672827 was granted and assigned to MindMeld on June, 2017 by the United States Patent and Trademark Office.
A conversation model is generated based on a conversation between a plurality of participants. Conversation text associated with the conversation is retrieved. A plurality of conversation model components are identified within the conversation text. A correlation score is determined for each pair of conversation model components representing a measure of relatedness between the pair of conversation model components. Extrapolated conversation model components are identified based on the plurality of conversation model components and the determined correlation scores. The conversation model components, the additional conversation model components, and the correlation scores are then stored as a conversation model.