Sentiment Extraction from Unstructured Text using Tabu Search-Enhanced Markov Blanket

摘要: Extracting sentiments from unstructured text has emerged as an important problem in many disciplines. An accurate method would enable us, for example, to mine on-line opinions the Internet and learn customers’ preferences economic or marketing research, leveraging a strategic advantage. In this paper, we propose two-stage Bayesian algorithm that is able capture dependencies among words, and, at same time, finds vocabulary efficient purpose of extracting sentiments. Experimental results on Movie Reviews data set show our select parsimonious feature with substantially fewer predictor variables than full leads better predictions about sentiment orientations several state-of-the-art machine learning methods. Our findings suggest are captured by conditional dependence relations rather keywords high-frequency words.