Emotions and Monoamines: New Approach to the Emotional Text Classification in Sentiment Analysis
The paper presents the classification of Internet-texts in Russian according to their monoamine status. We consider the levels of serotonin, noradrenaline and dopamine supposedly present in the blood of the text producer. The procedure of the non-discrete annotation of emotional tonality was used for the corpora data to identify the verbal markers for low-serotonergic, high-serotonergic, low-noradrenergic, high-noradrenergic, low-dopaminergic and high-dopaminergic texts. It is based on the Lövheim Cube model which includes eight emotional classes. The assessment results were mapped to a particular point in the Cube’s 3D-space. We identified six subcorpora from the assessed data according to their monoamine status. The parameters of verbal structures proper to the corpora-antipodes provided by Sketch Engine gave us potential serotonergic features such as absolutist words, adjective-noun combinations, numerals, noradrenergic features – indefinite and demonstrative pronouns, quantifier words – and dopaminergic features – combinatorics of кaк, etc. The obtained results give a perspective for the sentiment analysis which relates not only the linguistic representation with its corresponding emotional status but an individual’s biochemical response as well, thereby representing the wider mechanism of emotion generation.