English metaphor and emotion recognition based on cognitive function

In this paper, Bi-LSTM is taken as the core, and CNN and one-way LSTM are combined to construct a multi-level English metaphor recognition model. Based on the idea of multi-feature synergy, dependency feature, semantic feature and part of speech feature are used to enrich the learning information of the model. Meanwhile, the random walk model is used to determine the emotional labels of candidate words, where a metaphorical emotion recognition method integrating emotional scenes is constructed. The experimental results verify that the model can improve the performance of different levels of metaphor recognition tasks.