产品舆情环境形象评估方法构建与应用

Construction and application of product public opinion environmental image evaluation system

摘要:

按照人们对产品健康影响、环境污染方面的情感认知，开发了用于评估产品环境形象的人工智能方法。在建立产品环境形象的维度和情感极性分析框架基础上，以产品的互联网传播大数据为信息源，人工标注分类语料，采用自然语言处理技术和卷积神经网络CNN算法，训练验证了分类模型。以33种高风险和高污染产品为分析对象，将获取的互联网上与产品相关的新闻、评论或公众言论进行模型分类和环境形象评估。结果表明，环境舆情判定模型F1值为0.91，舆情分类模型F1值为0.45。产品的环境情感极性均以正面情绪为主，受舆情讨论热点程度影响，舆情数量多的产品情感倾向占比趋于均衡化，舆情数量少的产品情感倾向占比趋于极端化。化妆品、药品的环境健康风险最高，其中化妆品类的高环境风险产品是凡士林和角鲨烯，主要形象特征是过敏和生物有害性；药品类的高风险产品是咖啡因，主要形象特征是上瘾和生物有害性。

Abstract:

An artificial intelligence method according to people's emotional cognition of product health impact and environmental pollution for evaluating product environmental image has been developed. Based on the analysis framework of dimensions and emotional polarity of product environment image, the classification model is trained and verified by using natural language processing technology and convolutional neural network (CNN) algorithm, with big data of product spread on the Internet as the information source and manual annotation of classification corpus. Taking 33 kinds of high-risk and high-polluting products as the analysis object, the product-related news, comments or public comments on the Internet are classified and environmental image evaluation. The results show that the F1 value of the environmental public opinion judgment model is 0.91, and the public opinion classification model is 0.45. The environmental emotional polarity of products is dominated by positive emotions, affected by the degree of public opinion discussion hotspots, the proportion of emotional tendency of products with more public opinions tends to be balanced, which with less public opinions tends to be extreme. Cosmetics and drugs have the highest environmental health risk, among them vaseline and squalene are the high environmental risk products of cosmetics, and the main image characteristics are allergy and biological harmfulness; The high-risk product of drugs is caffeine, which is characterized by addiction and biological harmfulness.