基于蒙古语情感分布学习的数据增强方法

A Data Enhancement Method Based on Mongolian Emotion Distribution Learning

现有的情感分布学习尚未应用于蒙古语中,并且暂无利用情感分布学习进行数据增强研究。基于此,将极坐标的思想融入普鲁契克情感轮中,提出一种极坐标情绪表示法,把情感分布转化为复合情绪向量并将情感轮注意力信息融入模型中进行蒙古语情感分布学习。利用普鲁契克情感轮中任意两种基本情绪可以混合构成二元情绪的特性,为预测复合情绪向量扩展更为丰富的情绪标签,从而达到扩充数据集的目的。在蒙古语和中英文数据集上进行对比实验表明,基于极坐标情绪表示法的情感分布学习的性能优于传统方法。

The existing emotion distribution learning has not been applied to Mongolian,and there is no research on the use of emotion distribution learning for data enhancement.Based on this,this study integrates the idea of polar coordinates into the Plutchik's wheel of emotions,proposes a polar coordinates emotion representation method,and transforms the emotion distribution into com pound emotion vector and integrates the atte ntion information of the emotion wheel into the model for Mongolian emotion distribution learning.By using the featu re that any two basic emotions in Plutchik's wheel of emotions can be mixed to form binary emotions,we expand more rich emotion labels for the predicting composite emotion vector,so as to achieve the purpose of expanding the dataset.Comparative experiments on Mongolian,Chinese and English datasets show that the performance of emotion distribution learning based on polar coordinate emotion representation is better than that of traditional methods.