International Research Journal of Education and Technology

We endorse a collaborative multi-Trends sentiment type technique to teach sentiment classifiers for a couple of tweets simultaneously. In our technique, the sentiment statistics in exceptional tweets is shared to teach greater correct and robust sentiment classifiers for every Trends whilst categorized information is scarce. Specifically, we decompose the sentiment classifier of every Trend into components, a international one and a Trends-unique one.The international version can seize the overall sentiment information and is shared with the aid of using numerous tweets. The Trends-particular Greedy & Dynamic Blocking Algorithms model Naive Bayes and Drimux SVM can seize the particular sentiment expressions in every Trend. In addition, we extract Trends-particular sentiment information from each labelled and unlabelled samples in every Trend and use it to decorate the learning of Trends-particular sentiment classifiers. Besides, we include the similarities among tweets into our technique as regularization over the Trends-particular sentiment classifiers to inspire the sharing of sentiment statistics among comparable tweets. Two types of Trends similarity measures are explored, one primarily based totally on text and the opposite one primarily based totally on sentiment expressions. Moreover, we introduce green algorithms to clear up the version of our technique. Experimental consequences on benchmark datasets display that our technique can correctly enhance the overall performance of multi-Trends sentiment category and significantly outperform baseline methods.