Constructing automatic domain-specific sentiment lexicon using KNN search via terms discrimination vectors

Abstract

ABSTRACTWeb textual data content is a viable source for decision-makers’ knowledge, so are text analytic applications. Sentiment analysis (SA) is one of text mining fields, in which text is analyzed to recognize text writer implied opinion. In this paper, a new approach had been presented for automatic Arabic language sentiment lexicon constructing. Popular KNN search algorithm is utilized for this objective. Cosine distance between seeds terms and corpus terms is employed in KNN search query. Generated lexicon terms are launched from sentiment seeds and seeds terms are augmented via Arabic-specific NLP-based algorithm, which is helped to enhance seeds terms selection process. Term discrimination vector (TDV) is the main part of KNN query inputs TDV components are computed for each corpus term and it is constituted by four term weight techniques. According to the experimental results, TDV accomplished better results than TF-IDF traditional method with lower computation cost. Also, constructed lexicons outperformed premade lexicons accuracy results.