Towards an Analyzer of Emotions for Texts in Russian in Bilingual Perspective

This chapter presents the results of the experiment with two non-discrete annotations of the same sample of texts in Russian done, first, by Russian monolinguals and, second, by Tuvan-Russian bilinguals. Using statistical measures, we compared values obtained on four scales of emotional assessment interface in annotation done by 65 bilinguals and 174 monolinguals. As result, we saw statistically relevant differences on “shame–excitement” and “enjoyment–distress” scales: Bilinguals seem to be more sensitive to emotions of shame and distress than monolinguals and they show more significant value of emotion assessment magnitude parameter. Moreover, even inside the group of bilinguals, we discriminate statistically relevant discrepancies between balanced bilinguals and bilingual with dominating Russian language. Our main conclusion is that the concept of target-group of any emotion in text analyzer/classifier becomes necessary for modern projects in emotional text analysis.