Opinion has a strong influence on how people behave and therefore is a central part of all human activity. Before making an important decision, most people ask for an opinion of their closest friends and family, but big development of the Internet changed the way people express and search opinions. Nowadays, there are a lot of web pages where a person can express their opionon: social networks, portals, blogs, forums, etc. However, due to large amounts of available data, certain issues may arise. In the first place, there are a lot of different sources of data so people may have difficulties finding the ones that are relevant. Therefore, there is an increasing need for automatic detection of opinion, respectively sentiment analysis. This type of analysis is becoming more popular in many domains: politics, healthcare, production of consumer products, and various other services, etc. One of the tasks of sentiment analysis is classfication according to the polarity, respectively separation of texts in three categories: positive, negative and neutral. This paper consists of theoretical part where main concepts are explained and research part which is based on the analysis of the comments collected from portals „Index“ and „Jutarnji list“ and microblogs, that contain opinions about refugees. The analysis is divided into two parts: the first one contains frequency analysis of words in the sentence and results from a program which is using the prepared corpus for learning and testing the classifiers. Results from the analysis show that the most commonly used words in the comments are: „bravo“, „čovjek“, „izbjeglica“ and „eu“, and that the classifier with the highest percentage of accuracy for this corpus is maximum entropy. The second part of the analysis consists of the comparison of the used vocabulary and overall sentiment from two sets of data collected in two different time periods. Results are shown in the form of histograms and tables, and the program is written in programming language Python. The aim of the paper is to analyze the data and figure out what kind of opinion do people have about refugees, and to see if there is a difference in the opinion over a certain period of time.