The use of topic identification in opinion classification

With the development of the Internet and the increasing amount of opinion-based data, an ordinary person is not able to process them all. Therefore the importance of automated processing of this data grows. Processing of text data is very difficult. The presented approach is designed to help with the classification of posts in which the author describes his opinion. We focused on the possibility of increasing the accuracy of a return by identifying themes in contributions. In our work we achieved the accuracy 65.9 %. The precision was 60.5 % for positive and 79.2 % for negative reviews. On the other side the recall was 87.8 % for positive and 44.8 % for negative reviews.