"Detecting and visualizing changes in emotions in texts"
is a four-year research project (2010-2013) at the School of Computing, University of Eastern Finland. The project is funded by the Academy of Finland.
The project develops methods and tools for automatically detecting the writers' feelings and opinion towards certain topics. Particular interest in this project is the detection of emotional changes and visualization of the analysis results.

What emotions and opinions are and why it is important to automatically detect them?

An emotion is caused by a person, consciously or unconsciously, evaluating an event. Psychologists refer to this as appraisal.

Automatic detection and classification of sentiments has several potential areas of application. Marketing research, for instance, can benefit from using such tools for finding out what the general public thinks of their products. They can be used for monitoring newsgroups and forums, and for analyzing customer feedback.

More...

How can they be automatically detected?

The term sentiment analysis refers to this type of research. There are several ways to classify the sentiments present in a document. Polarity determines if the sentiment is positive or negative and subjectivity status refers to the distinction between subjective and objective statements (Pang & Lee, 2005). Attitude refers to the type of appraisal being expressed in a text as either affect ("angry", "sad", "bored"), appreciation ("beautiful", "fat", "tall"), or social judgment ("intelligent", "coward", "funny") (Whitelaw et al, 2005). Of particular interest in this project is what Pang & Lee (2005) refer to as agreement detection. The aim is to decide, given a pair of text passages, whether they contain the same sentiment polarities or attitudes towards a certain topic.

Although some isolated works, such as (Hearst, 1992) and (Wiebe et al., 1999), had already appeared, it was as late as 2001 when sentiment analysis caught a wider interest in the research community (Pang & Lee, 2005). While the research area is relatively new, it is based on various existing methods from Natural Language Processing (NLP), Text Mining, and Computational Linguistics. These methods are based on statistical and semantic analysis of texts. For example, self-organizing maps, various Information Retrieval (IR) methods, text or web mining, and information visualization can all be used for automatically determining a writer's attitudes and emotions with respect to the topic of a text.

More...