Sarcasm is notoriously tricky to convey through text, and the subtle changes in tone that convey sarcasm often confuse computer algorithms as well, limiting virtual assistants and content analysis tools. So researchers have now developed a multimodal algorithm for improved sarcasm detection that examines multiple aspects of audio recordings for increased accuracy. They used two complementary approaches — sentiment analysis using text and emotion recognition using audio — for a more complete picture.
