图文融合在线民宿评论情感分析模型

Sentiment Analysis Model for Online Home-stay Reviews Based on Integration of Image and Text

提出了一种基于图文融合的在线民宿评论情感分析方法,该方法对于情感分析中的文本信息,利用Word2vec构建主题聚类模型,通过主题中心词找到对应的主题属性字典,采用贝叶斯分类器进行情感分析,对比支持向量机(Support Vector Machine,SVM)、决策树方法,做出效果评估;对于情感分析中的图片信息,在卷积神经网络的基础上,训练图片分类模型;采用决策融合方法,计算图文情感概率,判断情感极性,将结果与用户实际打分作比较,有效地避免了用户打分与评论中表达情感不一致的问题。试验结果表明,图文融合的情感分析方法对图文评论表现出更好的情感分类效果。

This paper proposes a sentiment analysis method for online home-stay reviews based on the integration of image and text.It makes a sentiment analysis of the text information in the sentiment analysis,using Word2vec to construct a topic clustering model and find the corresponding topic attribute dictionary through the topic headword.In addition,it uses Bayesian classifier for sentiment analysis,compares SVM and decision tree methods and makes an effect evaluation.For the image information in sentiment analysis,the image classification model is trained on the basis of convolutional neural network.Finally,the decision fusion method is adopted to calculate the emotional probability of image-text,judge the emotional polarity,and compare it with the user’s actual score,which effectively avoids the problem that the user’s score is inconsistent with the emotion expressed in the comments.The experimental results show that the sentiment analysis method of image-text fusion has better emotional classification effect on image-text reviews.