Analysis of three methods for web-based opinion mining

Abstract

For the purpose of measuring semantic orientation of documents, we implemented an opinion mining tool which hybrids three different methods: The first one is based on semantic patterns, which simplify the structure of the natural language syntax; the second is based on the weighted sentiment lexicon, which used as semantic feature words; and the third one is based on traditional KNN or SVM text classification method. Our experiments show that each method has its own shorts and advantages.