Extraction and classification of rhetorical sentences of experimental technical paper based on section class

An extraction process of rhetorical sentences has become one of the problems in the study of automatic text summarization with a rhetorical sentence basis. Rhetorical sentence with high accuracy will be needed for producing a good summary. To improve the accuracy, this paper proposes a method in how to extract rhetorical sentences from experimental papers according to their category. The four main categories of experimental papers include problem, data, method, and result. Moreover, this paper proposes a section class as feature. We also calculated the frequency occurrence of rhetorical sentence in every section class. In our evaluation, we used tree algorithms including Naive Bayes, SVM and Decision Tree. Generally the SVM algorithm is proven to be better than the two other algorithms because the difference in value of the section class and non-section class feature is more reasonable. Overall, the rhetorical sentence extraction using section class has a better performance compared to those without class section.