Sentiment Analysis Tool using Machine Learning

This paper first presents the task of sentiment analysis and discusses its pros and cons when performed on data extracted from the web. Next, the problems of sentiment analysis in the Czech language are introduced. Currently, there is no sentiment dictionary available for the Czech language. To cope with this problem, supervised machine learning techniques are presented in the article as an approach for the sentiment analysis of Czech web pages. As part of this research, an application was developed using this approach for the analysis of web content based on language sentiment. An overview of its structure is included in the next part of the paper. The application modules cover automated text data mining from web pages, text processing and transformation, and machine learning based sentiment classification of the text. Several classification algorithms were applied and tested on the Czech text datasets, and their overall accuracy evaluated. The results of a comparison of the accuracy of the algorithms in the given area are also included in this paper. Keywords: sentiment analysis, web content mining, text processing, machine learning.