Classification of the Stance in Online Debates Using the Dependency Relations Feature
Thiri Kyaw, Sint Sint Aung · Global Society of Scientific Research and Researchers - International Journal of Computer · 2020
Online discussion forums offer Internet users a medium for discussions about current political debates.The debate is a system of claims regarding interactivity and representation.Users make claims to support their position in an online discussion with superior content.Factual accuracy and emotional appeal are critical attributes used to convince readers.A key challenge in debate forums is to identify the participants' stance, each of which is inter-dependent and inter-connected.The proposed system takes the post's linguistic features as input and outputs predictions for each post's stance label.Three types of features including Lexical, Dependency, and Morphology are used to detect the post's stance.Lexical features such as cue words are employed as surface features, and deep features include dependency and morphology features.Multinomial Naïve Bayes classifier is used to build a model for classifying stance and the Chi-Square method is used to select the good feature set.The performance of the stance classification system is evaluated in terms of accuracy.By analyzing the surface and deep features capturing the content of a post, the result of stance labels for this proposed system represents as for and against.