Classification of opinions in conversational content

Abstract

Nowadays, with enhancing possibilities of the Internet usage, the number of its users grows as well. People use it more and more to communicate among themselves. This kind of communication plays a significant role in the decision-making process. Based on this finding, a need to analyze the content of the ample web discussions (so-called conversational content) using the computers arose. Therefore, the following article deals especially with the issue of opinion analysis, more specifically the classification of opinions. We have created an algorithm, which allows determining the polarity of the text. With the analysis of text we can also process the intensification, negation and their combinations. We have created 4 classification dictionaries divided according to the types of words they contain. We have subsequently tested the algorithm with average accuracy of 86%.

Combined Approach for Sentiment Analysis in Slovak Using a Dictionary Annotated by Particle Swarm Optimization

“…Manually created dictionaries separate words into positive and negative groups [14] or provide also additional lists of words such as shifters (words that can change polarity) [20], [10]. Warriner lexicon [25] or Mikula lexicon [13] provide a value of polarity to each word. Automatically generated dictionaries require less human effort.…”