基于时间序列的方面级网络舆情情感演化模型

Aspect-level network public opinion sentiment evolution model based on time series

针对网络舆情情感分析主题词抽取不精确和文本静态化分析问题,论文提出了一种基于时间序列的方面级网络舆情动态情感演化模型ARMA-ALEE。通过方面级情感分类模型获取方面词和情感极性值,并对方面词使用过滤算法优化,再通过困惑度和JS散度确定最终方面词个数,进一步地还基于ARMA时间序列模型对方面词、方面词强度和方面词相关性的ARMA-ALEE模型动态地进行网络舆情情感演化分析。实验表明,该模型的情感演化研究取得了较好的结果。

Aiming at the problems of inaccurate subject word extraction and static analysis of texts in sentiment analysis of network public opinion, this paper proposes an aspect-level dynamic sentiment evolution model ARMA-ALEE of network public opinion based on time series. The aspect words and sentiment polarity values are obtained through the aspect-level sentiment classification model, and the filter algorithm is used to optimize the aspect words, after fhat the final number of aspect words is determined by the perplexity and JS divergence. Thereafter the evolution of network public opinion sentiment is dynamically analyzed for aspect words, the strength of aspect words and ARMA-ALEE model related to aspect words based on ARMA. Experiments show that the emotional evolution research of this model has achieved good results.