基于模型融合的电影短评情感倾向分析
首发时间:2017-02-08
摘要:为充分兼顾不同模型的特点,本文提出了一种基于模型融合的互联网电影短评情感倾向分析方法。本文以互联网电影短评为研究对象,首先从词汇层面入手,借助种子情感词汇,基于情感词典拓展的方式计算不同电影短评的情感倾向;随后从文本层面,引入CNN(Convolutional Neural Network,卷积神经网络)模型获取文本特征向量并得到电影短评的情感倾向。最后融合词汇和文本层面的情感倾向计算结果,生成电影短评的情感倾向。经过实验评估表明,本文提出的模型融合方法在情感倾向计算的准确率上比原有模型提高8%,利用模型融合的方法对电影短评进行情感倾向分析是有效的。
For information in English, please click here
Analysis of emotional tendency in film criticism based on model
Abstract:In order to fully balance the characteristics of different models, this paper presents a model based on the fusion of the Internet film criticism emotional tendencies analysis. This paper takes short film user comments as the research object, and starts from the lexical level, with the help of seed emotion words, based on the expansion of emotional dictionary to calculate the sentiment tendencies of different films' comments. Then, this paper introduces CNN (Convolutional Neural Network) model from the text level to get the feature vector of the text and get the sentiment tendency. Finally, this paper combines the results of vocabulary-level and text-level methods to generate the overall sentiment tendency of the film's short comment. Experiment evaluation result shows that the model fusion method proposed in this paper can improve the accuracy of calculating sentimental tendency by 8% compared with the original model. So it is effective to use the model fusion method to analyze the sentiment tendency of the film's short comments.
Keywords: Model Fusion Emotional Dictionary CNN Model Emotional Tendency
基金:
论文图表:
引用
No.4718051118133214****
同行评议
共计0人参与
勘误表
基于模型融合的电影短评情感倾向分析
评论
全部评论0/1000