A Quantitative Method of Text Emotiveness Evaluation on Base of the Psycholinguistic Markers Founded on Morphological Features

摘要

A new quantitative approach to identifying emotionally colored texts that reflect the excited state of its authors is proposed. This approach uses special psycholinguistic markers of text based on morphological characters ratios of Russian language. To apply such markers the morphological parser in combination with an ensemble of SVM classifiers was developed. Testing results of selected topics texts are presented. The study was conducted using texts from different sources such as news, regular blogs, microblogs, social network posts. It showed that results of developed approach can be a useful extension of Big Data methods of traditional sentiment analysis and it can be applicable to developing methods of personality computing.