Opinion Lexicon Automatic Construction on Arabic language

Abstract

The most important web content is textual data, especially for decision makers in many domains. Text mining demands a significant focus in opinion mining or Sentiment analysis (SA), wherein text is analysed to understand text writer implied opinions. In this paper, we present a new approach for automatic opinion lexicon constructing on Arabic language. Since opinion lexicon is used for lexicon-based sentiment analysis approach, popular KNN search algorithm is inquired via some features to choose lexicon's terms. Constructed lexicon contents are gathered, initiating from a few number of terms seeds. And a term discrimination vector (TDV) is composed for each corpus term. The experimental results are very promising compared to other opinion lexicons.