基于神经网络方法的隐喻识别研究综述

Review of Metaphor Detection Research Based on Neural Network Method

隐喻在日常交流和文学作品中十分常见,文本中隐喻表达的识别,是自然语言处理中的一个重要任务。隐喻识别任务中所采用的方法十分广泛,从基于规则的方法,利用语法特征,到传统机器学习算法,人工提取特征,再到近几年的神经网络算法,自动学习特征。其中,神经网络方法在隐喻识别中起到重要作用。对近些年应用于隐喻识别任务中的神经网络方法进行回顾与总结,并介绍常用隐喻识别数据集。

Metaphor is very common in daily communication and literary works.The detection of metaphor in text is an important task in natural lan guage processing.There are a wide range of methods used in metaphor detection task,from rule-based methods which use grammatical fea tures,to traditional machine learning algorithms which manually extract features,to neural network algorithms in recent years which auto matic learn features.Neural network plays an important role in metaphor detection.Reviews and summarizes the neural network methods applied to metaphor detection task in recent years,and introduces the commonly used metaphor detection corpus.