초록 열기/닫기 버튼
Recently, there is an increasing demand for the analysis of mass opinions using online text data. In particular, many studies have focused on automatic recognition of the main idea of subjective, argumentative writing. Additionally, such automatization of this task is fast becoming indispensable. This study constructed text data using debates in Korean on certain political issues, and attempted to identify the stance that each text supports about a given topic. We collected words which support one stance over the other and used them as the features for a machine learning classifier with a dictionary of sentiment words annotated based on their polarities. We then calculated weights for each sub-unit in the text based on the relevant discourse relations. Our classifier resulted in a slight improvement with respect to the defined weights.
