Identifying User Attitude Using Sentiment Analysis in Social Media

Abstract

Natural language is universal, flexible, and widespread but can’t exist without ambiguity. The ambiguity of a word or phrase contains in its having more than one attitude in the language to which the word belongs. The context in which an ambiguous word is used often makes it evident which of the emotion intended. This chapter presented emotion prediction model implementation. Firstly, the annotated tweets datasets and lexicons were defined. Secondly, preprocessing and feature extraction were discussed. The preprocessed tokens are transformed to semantic representative feature vectors. These vectors are trained using multiple machine learning techniques to build predictable emotion detection models. Real tweets datasets are used to train and evaluate the proposed emotion prediction systems Various snapshots and evaluation metrics from python and scikit learn packages are displayed to illustrate learning process and evaluation results during both the training and testing phases. Finally, discussion is presented to show the proposed emotion prediction models conclusions and contributions with respect to state-of-art researches in this field of study.