The Temperature of Feedback Chatbots: an Experiment in Humor Detection

Abstract

Humor is one of the measures of human intelligence, and similarly, humor is a major indicator of chatbot intelligence. To make chatbot answers more lively and interesting, thus increasing the interaction and entertainment between users and bots and providing a more interesting user experience. This paper collects a large text dataset and use models such as humor detection and humor scoring to detect the humor ability of chatbots, and improve the future work based on the results obtained from the experiments. It is shown that the CNN model's ability to recognize the humorous elements in the text and score the humorous information in the chat content gradually improves with multiple detections. This experiment implements the detection of humor ability of chatbots. A rating mechanism can assist in enhancing the humor and interactivity of robots, improve the existing computer humor model, enhance computer understanding of natural language, and promote the further development of artificial intelligence.