Improving Sentiment Polarity Identification on Twitter Using Metaclassifiers

Abstract

The exponential growth of social networking platforms has led many researchers to focus on ways of mining information from them. In this paper, we will use texts from social media in conjunction with techniques of Natural Language Processing to design a system that helps business organizations to identify polarity indicators from customer feedback. In this paper, we analyze tweets related to perceptions of an airline company, and detect the polarity of such tweets, using preprocessing and processing techniques common to the area, and to later incorporate the same techniques, in  a new methodology that consists of the incorporation of lexical resources (LR) and metaclassifiers to support the said task, thereby achieving a decision system with greater precision. In the present work, relevant results are reported in the area of NLP, making use of pre-processing and processing techniques known within the area, the main idea is to find the best classification scenario and increase the classification precision, for this the incorporation of lexical and metaclassifier resources was carried out.