Microblogging Sentiment Analysis Method With Combination Of Semantic Rules And Emoticon Weithting

Microblog has become the indispensable social media in people life since the Wenzhou crash event. The analysis of microblog sentiment can be fully used in many aspects such as product evaluation and social public opinion.In current microblog sentiment analysis field, researchers tend to classify the polarity of microblog sentiment, which is mainly used two methods: one is based on sentimental knowledge, including emotional dictionary, semantic rules, etc; the other is based on characteristic classification, including machine learning and others. For each method,considering as many influential factors as possible is the key to improve the accuracy of algorithm.For improving the accuracy of sentimental analysis, this article try to take more factors into consideration. Firstly, consider emoticon as an important element and calculate the sentiment of microblog with text in a weighted way, thus making an improvement in analysis of microblogs contain emoticon. Secondly, on semantic rules, I include several most common sentence rules and sentence relationship rules, such as exclamatory sentence, rhetorical questions, turning relationship and progressive relationship, to make a better result of sentimental analysis of complex statements, and the weights are determined by experiment. In addition, through the word similarity computing, calculate the sentiment of non-sentimental sentences and deliver a quantitative value to each blog.Besides accuracy, recall rate and F value, I propose another indicator named qualified rate, to measure the accuracy of each quantitative value.Finally, through Java program, 10000 sets of data were tested and verified the validity of the fusion algorithm.