US Patent 11770585 Reaction based real time natural language content adjustment

Using a sentiment analysis model, a reaction to a presentation of a first content is measured. Using the first content and the reaction, a content generation parameter is determined. Using a first content generation model and the content generation parameter, second content corresponding to the first content is generated. Using a second content generation model and the content generation parameter, third content corresponding to the first content is generated. The first content is adjusted by combining the first content with the second content and the third content, the adjusting resulting in an adjusted content.