Abstract: Online product reviews have become the primary means to enable people to explain their own views on a particular commodity. And, the research on the sentiment analysis model owns values in both business and academic areas. Discussing on several machine learning models for sentiments analysis, using enlarged emotional dictionaries, and describing full machine learning procedures, this paper proposes a set of sentiment analysis model for the sentiment analysis on the catering industry. Then, this paper discusses some classify algorithms, such as Naive Bayes and C45, and gives detailed discussions about effects of different models based on various evaluation methods. The experimental results show that the proposed model gives full play to emotion dictionary efficiency, and is more suited to judge customer emotional tendencies.

Key words: online product reviews, sentiment analysis, emotion dictionary, machine learning, model evaluation

摘要： 在线商品评论已成为对商品阐述看法的主要手段，对商品评论的情感分析研究具有学术及商业价值.研究情感分析领域若干机器学习模型，通过扩充情感词典，运用机器学习方法，设计餐饮领域网上评论情感分析模型.深入探讨朴素贝叶斯、C4.5等分类算法，利用多种性能评价方法，详细讨论不同模型的分析效果，结果表明所设计模型发挥出情感词典的有效性，更加适合于判断客户情感倾向.

关键词: 网络商品评论, 情感分析, 情感词典, 机器学习, 模型评价