Comparison of Probabilistic Corpus Based method and Vector Space Model for Emotion Recognition from Poems

摘要

This paper discusses the automatic detection of emotions in English poems. Emotions are classified based on ?Navarasa? which is described in ?Natyasastra?. Navarasa consists of nine basic emotions such as Love, Sad, Anger, Hate, Fear, Surprise, Courage, Joy and Peace. We have manually created an emotions tagged corpus from poems. Using poems mined from the web, we applied corpus-based tagging method to recognize the emotion of a poem. For the emotion recognition, we have used the Vector Space Model with a total of 348 poems of 165 poets mined from the web. We have approached this problem from four perspectives.Traditional Vector Space model Vector space model, eliminating stop words (emotionless words) and without stemming Vector Space model without eliminating stop words and without stemming. Corpus-Based emotion recognition system. The Traditional Vector Space model gives better performance than other methods.