Short Text Similarity Measures for Social Sentiment Comments

Abstract

The feeling of social networks is the attitude and feelings people have about their brand on social networks. Adding context to all commissions, comments, and actions need to analyze. It is important that the brands listen carefully to what is said about their online business. And more importantly to know whether the conversation is positive or negative. In this paper, we find the similar groups of comments which talk about the same context depending on particular topic. By doing so, we can help online business works to group the customers who shares common interest and same feeling of their products. This paper introduces how to find the similar text in semantic ways in both word-level and phrase-level measures by filling the gap of syntactic measures in text similarity. For the datasets, Twitter dataset is used for system implementation because their comments are short and compatible with our proposed system. According to the experimental results, the results get promising results in terms of higher accuracy rate but lower error rate by switching two datasets available from online.