Argumentation mining: classifying argumentation components with Partial Tree Kernel and Support Vector Machine for constituent trees on imbalanced persuasive essay

Abstract

Argumentation Mining is a method that automatically identified an argument structure in a text document. The structure of this argument consists of several components that become very important to evaluate itself. This study builds a classification system of 3 classes of argumentation components on persuasive essays so that the data is multi-class using Tree Kernel which is part of the pre-processing and Support Vector Machine as a tool for grouping this text. This research is an adaptation of research conducted by Lippi and Torroni where they used 2 classes to get 74.6% precision, recall 68.4% and F1- score 71.4%, while this study used 3 classes of argument class and managed to get a precision value of 79%, recall 78% and F1-score 74% by using the Sampling Method to overcome the problem of the amount of data imbalance.