面向社交媒体的幽默计算理论与实践

Theory and Practice of Humor Computing for Social Media

幽默是人类独有的语言表达方式,具有幽默感的人可以构建和谐的交际氛围。如何识别社交媒体上的海量幽默文本和评价用户幽默感成为计算语言学领域面临的重大挑战。因此,我们从认知语言学的视角,提出从资源建设、幽默识别到幽默感评价的社交媒体幽默计算框架。首先,参照幽默的语义脚本理论,构建面向社交媒体的中文幽默语料库。然后,利用情感倾向性、语义不一致性及模糊性特征,对文本进行幽默识别。最后,从创作和欣赏两个维度进行用户的幽默感评价。幽默计算将有助于提升语言智能的水平,有助于提高阅读理解、人机对话、机器翻译等领域的应用性能。

Humor is a unique language expression of human beings,and people with a sense of humor can build a harmonious communication atmosphere.How to identify massive humor texts on social media and evaluate humor sense of users has become a major challenge in the field of computational linguistics.Therefore,from the perspective of cognitive linguistics,we propose a social media humor computing framework that covers resource construction,humor recognition,and humor sense evaluation.Firstly,in light of the semantic script theory of humor,we construct a Chinese humor corpus for social media.Then,based on semantic representation,combined with sentimental tendencies,semantic inconsistency,and fuzziness features,we recognize the humor on the text.Finally,based on obtaining the humor attributes and behavioral data of social media,we evaluate the user's sense of humor from the dimensions of creation and appreciation.Humor computing will help improve the level of language intelligence and enhance the application performance in areas such as reading comprehension,human-machine dialogue,and machine translation.