属性归类与条件句倾向性分析

发布时间：2018-04-22 03:14

本文选题：属性归类 + 条件句　；参考：《大连理工大学》2012年硕士论文

【摘要】：本文主要分析产品属性的情感倾向性。由于语言习惯不同,人们在写评论时对同一种产品属性会采用多种表达,需要对这些不同的表达进行归类,另外人们也会采用不同的句型,需要对特殊句型中属性的倾向性进行处理。本文的工作主要分为两个相连的部分：属性归类和条件句中属性的倾向性分析。本文对属性进行归类时,考虑到同类属性虽然有不同的描述,但是在句中却和相同的情感词搭配使用,因此,首先抽取评论句中属性和情感词的搭配关系,形成一个二部图,然后用权重标准化SimRank计算不同属性之间的相似度,并把所得的结果与半监督学习中的贝叶斯分类器进行融合,得到了更好的分类结果。本文分析条件句中的属性倾向性时,首先识别条件句,其次分析属性倾向性。条件句中一般都含有条件连接词,但是有些条件句中没有条件连接词,称为隐式条件句。经过观察,发现隐式条件句中含有一些体现条件关系的词,称之为隐式条件词。识别条件句时,主要依据条件连接词和隐式条件词及其词性以及分类序列规则进行分类；分析属性倾向性时,依据条件连接词和隐式条件词把条件句分为假设条件句、让步条件句、特定条件句和无条件句四类,并把条件句的类别用于SVM分类。通过实验证明了本文方法的有效性。
[Abstract]:This paper mainly analyzes the emotional orientation of product attributes. Because of the different language habits, people will use a variety of expressions for the same product attributes when writing comments, which need to be classified. In addition, people will also adopt different sentence patterns. We need to deal with the tendency of attributes in special sentence patterns. The work of this paper is divided into two parts: attribute categorization and attribute tendency analysis in conditional sentences. When classifying attributes, this paper takes into account that similar attributes are different in description, but they are used with the same affective words in sentences. Therefore, first of all, the collocation of attributes and affective words in comment sentences is extracted to form a bipartite graph. Then the similarity between different attributes is calculated by using weighted standardized SimRank, and the result is fused with the Bayesian classifier in semi-supervised learning to get a better classification result. In this paper, we first identify the conditional sentence and then analyze the attribute tendentiousness. Conditional sentences generally contain conditional connectives, but in some conditional sentences there are no conditional connectives, which are called implicit conditional sentences. After observation, it is found that there are some words in implicit conditional sentences, which are called implicit conditional words. In identifying conditional sentences, they are mainly classified according to conditional connectors, implicit conditional words and their parts of speech and classification sequence rules, and conditional sentences are classified into hypothetical conditional sentences according to conditional connectors and implicit conditional words when attribute tendency is analyzed. Concessional conditional sentence, specific conditional sentence and unconditional sentence are divided into four categories, and the category of conditional sentence is used in SVM classification. The effectiveness of this method is proved by experiments.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：H043

【参考文献】