Determining Sentiment of Tweets Using First Bosnian Lexicon and (AnA)-Affirmative and Non-affirmative Words

The paper presents a model for sentiment tagging of Twitter messages based on (AnA)—affirmative and non-affirmative words and tagged sentiment dictionary/lexicon. Lexicon was used in combination with intensifiers, where intensifier (AnAword), that stand next to the word, produce more powerful (more positive or more negative) sentiment value of word—or the whole tweet. The result of the study was creating the first Bosnian sentiment lexicon that can be used for sentiment analysis of the Bosnian language. Besides that, the paper presents a more powerful way to measure sentimental value by using AnAwords. By using confusion matrix model’s overall accuracy was counted for both positive and negative sentiment value. Good overall evaluation of model was presented with huge values 0.9075 and 0.9004 for positive and negative sentiment value, respectively.