基于特征权重计算方法的情感分析

发布时间：2023-01-30 18:55

　　近年来,情感分析一直是自然语言处理研究者群体日益关注的主题。情感分析可以帮助公司和公共管理部门的人员更多地了解客户的意见,并帮助他们做出一些决定。在本文中,我们首先介绍了情感分析任务的背景,定义,以便读者更好地理解本文的研究目标以及论文的贡献。我们还介绍了最近的几种情感分析的方法,如概率算法（朴素贝叶斯）,最近邻算法和变量算法,决策树或分类和矢量支持机器。然后介绍了构建情感分析系统的步骤,包括预处理,特征提取和性能评估。最后,我们更加关注由在线酒店评论组成的数据集,并应用监督机器学习方法Na?ve Bayes使用unigram特征和两种类型的信息（频率和TF-IDF）来实现文档的极性分类。如我们的实验结果所示,在准确性,精确度,召回率和F_score方面,我们的模型优于其他模型。 

【文章页数】：62 页

【学位级别】：硕士

【文章目录】：
摘要
Abstract
List of abreviations
Chapter Ⅰ Introduction
    1.1.Research Background
    1.2.Research Significance
    1.3.Research Status
    1.4.My Contribution
    1.5.Thesis Outline
Chapter Ⅱ Related Work
    2.1.Sentiment Analysis
        2.1.1.Definitions,Tasks,and Terminology
        2.1.2.Opinion Sentiment Analysis
    2.2.Sentimental Analysis Customer Review
        2.2.1.Sources for Online Reviews
        2.2.2.Formats of Online Reviews
        2.2.3.Specifics of the Application Domain
        2.2.4.Subtasks in Sentimental Analysis Customer Review
    2.3.Text classification approaches
        2.3.1 Probabilistic Algorithms(Naive Bayes)
        2.3.2 Algorithm of the Nearest Neighbor and Variants
        2.3.3 Decision Trees or Classification
        2.3.4 Vector Support Machines
Chapter Ⅲ Sentimental Analysis System Construction
    3.1.Approaches
    3.2.Preprocessing
    3.3.Feature extraction
    3.4.Evaluation Metrics
    3.5.Visualize Results
Chapter Ⅳ Methods and Results
    4.1 Methods
        4.1.1 Term frequency-inverse document frequency(tf-idf)
        4.1.2.Naive Bayes Algorithm
    4.2.Experimental Results
        4.2.1.Dataset
        4.2.2.Experimental Setting
        4.2.3.Evaluation Metric
        4.2.4.Experimental Results
        4.2.5.Results of Evaluation Metrics
Chapter Ⅴ Conclusions and Future Work
    5.1 Conclusions
    5.2 Future Work
攻硕士学位期间取得的研究成果
Acknowledgement
References
附件



本文编号：3733355