Demo for Commonsense Story Generation    Let our knowledge-enhanced pretraining model generate a reasonable story based on your beginning
Paper:Jian Guan, Fei Huang, Zhihao Zhao, Xiaoyan Zhu, Minlie Huang. A Knowledge-Enhanced Pretraining Model for Commonsense Story Generation. Transactions of the Association for Computational Linguistics 2020.[pdf][code] Our group:Conversational AI (CoAI) Contact info: j-guan19@mails.tsinghua.edu.cn
Instruction:
If you want to generate a story based on your beginning:
Choose k (e.g., 40) and temperature (e.g., 0.7) for sampling.
Enter an English sentence as the story beginning (e.g., [MALE] was cooking bacon.)
Then press the Get A Story button.
Wait for a while (about 10 seconds), see what happens.
Note that the system will automatically recognize names in your input and transform the male/female/neutral names into the special tokens [MALE]/[FEMALE]/[NEUTRAL], respectively. However, it is not so smart yet :), and we recommend you directly use the special tokens in your input.
If you want to share a better story:
Write down any story you want to share (without any restriction).
Then press the Share My Story button.
Sampling method     Top k sampling    
Temperature