Abstract:

In order to study the social emotion of ancient China, this paper puts forward a method of analyzing the emotion of metrical poetry based on the integrated capsule network by using the integrated capsule network model training through different rhythm with word count rule to discriminate the emotion of metrical poetry. It solves the difficult problem of the sentiment analysis of metrical poetry due to the implicit sentiment features and the too compact semantics, analyzes the emotion of metrical poems to the ancient periods, and mines the whole social emotion in each important period of ancient China. The experimental results show that the accuracy of this method is better than that of CNN and single capsule network, and the analysis results can reflect the social emotions of ancient China.

Key words: social emotion, emotion analysis, metrical poetry, computational sociology, capsule network

摘要：

为研究古代中国的社会情感，提出一种基于整合胶囊网络的格律诗情感分析方法，利用字数规则整合依据不同格律训练的胶囊网络模型对格律诗进行情感判别，解决格律诗文本由于情感特征不明显、语义过于紧凑导致的情感分析困难问题，通过分析格律诗情感对古代中国各个重要时期的社会整体情感状况进行挖掘。实验结果表明，该方法对格律诗情感分类的准确率优于CNN与单一胶囊网络，分析结果能够反映出古代中国的社会情感。

关键词: 社会情感, 情感分析, 格律诗, 计算社会学, 胶囊网络