Research On Chinese Micro-blog Opinion Sentence Identification Based On Three-Way Decisions

The number of microbloggers has been increasing sharply as the rapid development of the Internet and mobile Internet, and the importance of Micro-blog opinion sentence identification is growing for public opinion research. However, processing Micro-blog text manually is inefficient and time-consuming. Therefore, identifying Micro-blog opinion sentence automatically has been becoming one of fascinating areas of researches in recent years. Chinese Micro-blog opinion sentence identification based on three-way decisions is studied. Main contents of the research are as follows.Algorithms such as support vector machines, Naive Bayes, k-Nearest Neighbor algorithm are used in opinion sentence identification with good classification performance, high accuracy rate and re-call rate. Three-way decisions theory is introduced to construct three-way decisions classifier based on SVM, NB and KNN. The evaluation criteria for the three-way decisions classifiers and three-way decisions threshold algorithm based on expert knowledge are proposed. Experiment result shows that the proposed method can improve the identification accuracy effectively without sacrificing the classification performance. The validity of the three-way decisions method is verified.Three-way decisions threshold algorithm based on expert knowledge requires domain knowledge, which hinders wide application of the three-way decision methods. Therefore, an improved automatic learning three-way decisions threshold algorithm is proposed based on decision risk minimization principle. The proposed algorithm can learn optimal global threshold automatically without domain expert knowledge, and improves overall performance of the Micro-blog opinion sentence identification greatly.Although the three-way decisions threshold algorithm based on decision risk minimization principle can identify Micro-blog opinion sentence with high accuracy, the re-call rate and overall performance suffers. Therefore, A comprehensive performance evaluation model based on analytic hierarchy process classifier is presented. Combined with decision risk loss, three-way decisions optimal threshold model is build, and a three-way decisions adaptive threshold algorithm based on risk loss and performance evaluation is proposed. Experiment results show that the proposed algorithm can learn three-way decisions thresholds efficiently, and set weight and rate parameters flexibly to balance the risk loss and performance evaluation of the classifier. Therefore, improves the overall performance of the classifier effectively under acceptable decision risk loss.