Humorous dialogues are an important element of interpersonal communication and one of the important processes of human-computer communication towards intelligence. The purpose of this project is to develop related technologies and systems and to explore Chinese humorous dialogues so as to serve as a tool to enhance the dialogue experience, increase the satisfaction of application systems, stimulate creativity, and assist in interpersonal communication. In addition, because of its characteristics of subjectivity, geography, culture, current affairs, and language differences, humor has its own indispensable value in the study of Chinese humor dialogue. At the beginning of the project, the focus was on the collection and construction of a humorous corpus to facilitate subsequent research. The results in this regard were published in the international conference LREC, and a human-machine dialogue system was constructed on the Line platform to test its application effectiveness, which also has been written in a paper and published in a domestic journal. In the later stage of this project, an emotional conversation system is implemented using deep learning and other technologies such as GPT-2 and BERT, based on the corpus provided by the 2019 Chinese Emotional Conversation Generation (CECG) evaluation task. The effectiveness of the system is evaluated according to the test data and criteria provided by CECG. The results evaluated by three human annotators show that the system has a similar or even better effectiveness level with that of the best team participating in the 2019 CECG task. Further case studies reveal that the more post/reply pairs about a topic in the training data, the better the language model of GPT-2 to generate innovative, interesting, and perfect response sentences for that topic.
|Effective start/end date||2018/08/01 → 2020/07/31|
- Computational Humor
- Artificial Intelligence
- Chinese Humor Dialogue
- Humor Recognition and Generation
- Humor Corpus
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.