The Identification Of Emotion Metaphor Based On Field Conflict

Metaphor identification in text is a popular and open problem in natural language processing in recent years. As news report, network information widely spread through Microblogs, WeChat and other platforms, we can easily realize metaphoric expression in our daily life. According to statistics, every three sentences we use include one metaphor usage. It underlines the important role of metaphor in our Chinese language. Metaphoric expression usually reflect some emotion from the speaker or writer, thus it is so import and urgent to identify emotion metaphor in the field of Artificial Intelligence. Aiming at identifying metaphor and recognizing its emotion, helping computer learn and understand human emotions, then giving back a proper response is our target in this paper.Firstly, we introduce relevant technologies of building metaphor corpus to prepare an abundant knowledge base for metaphor identification and other machine translation tasks. During the building process of this corpus, we use crowdsourcing to annotate for sentences. Due to the complexity and the veiled meaning of metaphor, it is rather difficult to make a common standard to help people recognize and judge whether a sentence is metaphor or not. So we gave out a detailed explanation and representative samples. When collecting sentences, we filter the ambiguous ones and context-dependent usage. Besides, we use a tagging system to simplify annotation task and improve efficiency.After building the emotion metaphor corpus, we try two methods to identify metaphoric expression. Using annotated set of sentences, with machine learning methods including Support Vector Machine (SVM) and Long Short Term Memory (LSTM), we found it effective to classify metaphor usage from literal usage. Then we induce a metaphor identification model based on field conflict theory. Before this, we made a strict field rule in tree model and arrange every node a code. When finishing metaphor identification, the model can resolve emotion category for metaphoric expression. The outcomes of experiments verify its effectiveness. At last, we introduce some application area of metaphor identification.