文本幽默识别综述:从数据到方法

Review of Text Humor Recognition:From Data to Methods

幽默计算研究致力于利用计算机方法理解和识别幽默表达,挖掘幽默潜在的语义内涵,构建面向幽默的计算模型,实现幽默的自动识别和生成,提升人机交互智能程度.在开发基于幽默计算的人工智能系统的需求日益凸显的趋势下,通过文献调研方式进行幽默识别综述.首先,重点研究了幽默特征的提取方法;其次,从数据和方法两个维度总结了文本幽默识别的研究进展.归纳常用数据集的收集标注过程及特点,系统地对比了包括基于传统机器学习和基于深度学习的文本幽默识别方法;最后,对幽默识别领域的相关研究进行了总结与展望.

Humor computing research is devoted to understanding and identifying humorous expression,mining the potential semantics of humor,building the humor-oriented computing model,realizing the automatic recognition and generation of humor,so as to enhance the intelligence of human-computer interaction.With the increasing demand for the development of artificial intelligence system based on humor computing,this review is carried out by literature research.Firstly,it focuses on the extraction methods of humor features;Secondly,it summarizes the research progress of text humor recognition from the two dimensions including data and methods.This paper summarizes the collection and annotation process of common data sets,systematically compares the text humor recognition methods based on traditional machine learning and deep learning;Finally,some future prospects of humor recognition are proposed.