The Identification of the Emotionality of Metaphorical Expressions Based on a Manually Annotated Chinese Corpus

Abstract

Metaphorical expressions are frequently used to convey emotions in human communication. However, there is limited research on the detection of emotionality in metaphorical expressions, although a number of studies have focused on sentiment analysis and metaphor detection separately. We, therefore, attempt to identify emotions in Chinese metaphorical texts. We first construct a manual corpus with an annotation scheme, which contains annotations of metaphor, and emotional categories. We then use the corpus as a train-and-test set to identify the emotions in metaphorical expressions automatically with three methods. The first method is based on a field dictionary and field conflict. The second method is based on a support vector machine. The third method is based on deep learning, and it applies the long short-term memory model to identify the emotion of metaphor. The experimental results show that the third method performs better in identifying metaphor tasks, while the first method works better for emotion classification. In this paper, we compared the strength of heuristic, stochastic, and deep learning approaches, which contributes to a challenging natural language processing issue: the detection of emotionality in metaphor.