Mixed-sentiment classification of web forum posts using lexical and non-lexical features
Hikmat Ullah Khan · Journal of Web Engineering · 2017
The sentiment detection of the content has become an active research domain in recent years due to the increased availability of public views and opinions in the social web forums. Earlier works detect the sentiment arousal and valence using a lexicon or a dictionary. This paper aims to classify a post content in the social web forums by identifying the mixed-sentiment views and targets to find such posts in which users' views have both positive and negative emotions. Identification of the mixed-sentiment content has several potential applications such as monitoring public views, making products related business decisions and predicting users' behaviors. I propose a non-lexical feature set and compare with the conventional lexicon-based sentiment feature set. The four state of the art classification algorithms applied on the large dataset of public forum verify that the proposed non-lexical features are helpful to find the mixed-sentiment in online forums. The main contribution is the proposal and validation of such features which do not need a lexicon. In addition, a comprehensive analysis of the dataset has been carried out using the power law analysis. The features have been ranked according to their significance in the classification model to identify mixed-sentiment content in the social web forum.