Where Is Happily Ever After? A Study of Emotions and Locations in Russian Short Stories of 1900-1930
The paper tackles the problem of the automatic detection of emotions in literary texts using distributional semantics techniques. The experiment was carried out on the material of Russian short stories from the 1900-1930s. We investigated the emotional lexis distribution across different locations in narratives. At first, we calculated the semantic association score between each word in a corpus and a certain emotion label. This stage resulted in twelve ordered lists of semantic associates for each emotion label: fear, anger, sadness, happiness, shame, pride, disgust, pleasure, surprise, relief, amusement, and content. Then, we calculated the frequency of these emotional words per story and normalized the values. Anger turned out to be the most prominent emotion expressed in the texts. In the next stage, we compared obtained distributions with the information on where the action takes place in each story. That data was received from previous annotations by readers. Based on mean relative frequencies, the places with the highest emotionality appeared to be apartment, theater, Petersburg, pub, and hall. Additionally, the emotional diversity of a few thematically specific locations is considered.