Automatic Construction of Domain Specific Sentiment Lexicons for Hungarian

Sentiment analysis has become an actively researched area recently, which aims to detect positive and negative opinions in texts. A good indicator for the polarity of a given text is the number of words in it that have positive or negative meanings. The so called sentiment lexicons are lists containing words together with their polarities. In this paper we present methods for creating sentiment lexicons automatically. We use these lexicons in sentiment analysis tasks on general and domain-specific Hungarian corpora. We compare the efficiency of sentiment lexicons from different domains and show the importance of using domain-specific sentiment lexicons for different sentiment analysis tasks.