Conference Paper Automatic Narrative Humor Recognition Method Using Machine Learning and Semantic Similarity Based Punchline Detection

Rafal, Rzepka  ,  Yusuke, Amaya  ,  Motoki, Yatsu  ,  Kenji, Araki

Description
In this paper we introduce our method for recognizing jokes written in Japanese language by where the punchline is detected using WordNet. The results showed that when compared to method based on Bayesian posterior probability baseline, the pro- posed system achieved 5.3 point increase in recall and 2.6 point increase in classification accuracy. Our work1 is the first challenge to detect humor in Japanese language and this ability can be utilized not only for more natural reactions while perceiving user’s utterance, but also for discovering funny stories to be uttered by an agent.
Full-Text

http://eprints.lib.hokudai.ac.jp/dspace/bitstream/2115/63638/1/2015-A-3.pdf

Number of accesses :  

Other information