Journal of Digital Information Management Xiaona Jiang 15 4 2017 http://dline.info/fpaper/jdim/v15i4/jdimv15i4_5.pdf Metaphor is a result of emotion conceptualization hidden in human language. Due to the complexity and abstraction of human emotion, it is difficult to calculate and create a model for it. However, emotion modeling and calculation is of great significance in the process of machine translation (MT). Usually, incorporating the calculation of emotion metaphors in machine translation could make the language much more vivid and meet the standards of faithfulness, expressiveness and elegance in translation. Normally, calculation of emotion metaphor adopts machine learning and pattern identification, and it requires the samples from emotion metaphor corpus of large scale and high quality. The thesis builds an English and Chinese bilingual corpus with affluent emotion metaphors and supports data to emotion metaphor calculation by machine translation. In the process of building emotion metaphor corpus, 5 main procedures including theoretical framework, design principles, data collection, data annotation and index monitoring are illustrated. Finally, machine translation experiment has been done in emotion metaphor corpus built in this thesis, which adopts same recurrent neural network and LSTM mnemon to compare with existing machine translation corpora. Result shows that the emotion metaphor corpus built in this thesis is able to express emotion metaphor in machine translation.