A text mining approach to extract opinions from unstructured text

Abstract

Background/Objectives: To extract and interpret public opinion from informal description of text in social media websites. Methods: The informal descriptions containing opinions are Tokenized, Parts of Speech Tagging, Word-sense Disambiguation and Text Transformation/Attribute Generation are employed. Sample data pertaining to performance rating of cricket players was collected from twitter, cricinfo and cricbuzz. The reviews were rated against subjective evaluation criteria, the scale ranging from poor, moderate, good, excellent and the linguistic variables were converted into numerical values using fuzzy since the expressed opinions may be easy to understand. Findings: Opinions enhance any decision making process, so the influence of opinions from sample data pertaining to performance rating of cricket players are classified as poor, moderate, good and excellent. Application/Improvements: Using the proposed approach a generic prototype can be built that can be used to extract opinions and interpret.