Neural Metaphor Detecting with CNN-LSTM Model

• DOI: 10.18653/v1/w18-0913
• Publication Date: 2018-05-30T00:48:46Z
• Authors: Chuhan Wu; Fangzhao Wu; Yubo Chen; Sixing Wu; Zhigang Yuan; Yongfeng Huang
• Keywords: Softmax function

Metaphors are figurative languages widely used in daily life and literatures. It's an important task to detect the metaphors evoked by texts. Thus, metaphor shared is aimed extract from plain texts at word level. We propose use a CNN-LSTM model for this task. Our combines CNN LSTM layers utilize both local long-range contextual information identifying metaphorical information. In addition, we compare performance of softmax classifier conditional random field (CRF) sequential labeling also incorporated some additional features such as part speech (POS) tags cluster improve model. best achieved 65.06% F-score all POS testing subtask 67.15% verbs subtask.