摘要
目前的情感分析研究大部分仅局限于能够明显地表达意见的主观性文本,却没有对一些隐含地表达情感的文本进行分析。针对这一不足,提出一种基于条件随机场(CRFs)模型的意见挖掘中维吾尔语文本隐式情感分析方法。利用互信息(MI)衡量上下文的依赖度,结合词法、语境依赖词、标点符号和习语等特征用于隐式情感分析。在特征选择时,通过对信息增益(IG)进行改进,解决语料中数据集不平衡的问题。该方法用于维吾尔语文本隐式情感分析的准确率为77.11%,召回率为78.37%,表明了其在意见挖掘中隐式情感分析任务上的有效性。
Most of the current researches on the sentiment analysis can analyze the opinions that are expressed clearly in the subjective text, but fail to analyze the ones that are expressed implicitly. To solve this issue, a method based on the CRFs model was proposed to analyze the implicit sentiment of the Uyghur opinioned text. The mutual information （MI） was used to compute the dependence of the context, and the morphology, the word depending on context, the punctuation, idioms and other features were combined, which contributed to the implicit sentiment analysis. When these features were selected, the improved information gain fIG） was applied in the imbalanced dataset of the corpus. The experimental results show that the precision rate and the recall rate of the sentiment analysis reach 77.11% and 78. 37% respectively, which demonstrate the efficiency of the proposed method on the implicit sentiment analysis.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第9期3295-3300,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(61262064
60963017
61063026
61063043
61331011)
国家社科基金项目(10BTQ045
11XTQ007)
