Emotion level sentiment analysis: The affective opinion evaluation
Sentiment analysis evaluates writers' opinions based on pivot items extracted from text. These items are called opinion bearing words or, simply, sentiments. Based on these sentiments, sentiment analysis derives the opinion evaluation. Most of the work in this area evaluates opinions based on the polarity detection that can be positive, negative, or neutral. This coarse-grained sentiment polarity is insufficient to convey the precise affect of the writers. To overcome this limitation, this paper introduces emotions as a fine-grained alternative for sentiment evaluation. This can be realized through the use of a cognitive model of emotion representation that organizes the most commonly known emotions. The cognition model of emotion is mapped to an ontology. A semantic similarity is computed to measure the semantic relation between the given opinions and the emotions in the ontology. The mapping between rational and emotional sentiments is obtained by computing the correlations between them using the Google search engine. The initial results are promising.