CAJ | 학술논문

在研究区分性关键词提取方法的基础上，对维吾尔语中的生气和高兴等常见情感类型进行基于文本句子的情感分类研究。结合维吾尔文本句子中的情感表达特点，以词频和文档频率作为基本统计量，通过计算同一词语在不同组合统计量下的类间差异得到区分性关键词，并基于这些关键词进行特征提取和区分性情感模型构建。从维吾尔语电影字幕、小说等文本库中提取生气和高兴2种情感构造实验数据集，并验证所提出的情感分类方法。实验结果表明，基于区分性关键词的建模方法能有效地对维吾尔文本句子进行情感分类。
재연구구분성관건사제취방법적기출상，대유오이어중적생기화고흥등상견정감류형진행기우문본구자적정감분류연구。결합유오이문본구자중적정감표체특점，이사빈화문당빈솔작위기본통계량，통과계산동일사어재불동조합통계량하적류간차이득도구분성관건사，병기우저사관건사진행특정제취화구분성정감모형구건。종유오이어전영자막、소설등문본고중제취생기화고흥2충정감구조실험수거집，병험증소제출적정감분류방법。실험결과표명，기우구분성관건사적건모방법능유효지대유오이문본구자진행정감분류。
This paper presents a classification approach for Uyghur text sentiment,such as angry and happy,based on discriminative key word extraction. Combined with the characteristics of sentiment expression in Uyghur text, the term frequency and document frequency are derived as primary statistics. Various discriminative statistics which reflect the discrepancy of the positive and negative sentiment datasets are derived from the primary statistics for each vocabulary word, and are used to extract discriminative key words. Features are extracted based on these keywords and are used to train discriminative sentiment models. This paper builds a sentiment text database by excerpting two sentiments:angriness and happiness from Uyghur movie transcriptions and novels,and verifies the proposed approach. Experimental results show that the method based on discriminative keyword extraction is effective in Uyghur text sentence sentiment classification.