Abstract:

Abstract Sentiment Analysis comprises a variety of tasks, including subjectivity detection, polarity detection, sentiment magnitude detection, and emotion type recognition. The text is assessed as subjective in the subtask of subjectivity detection. Words and phrases' subjective nature vary based on their context; an objective text may include subjective observations. For example, news reports include quotations from people's viewpoints. In this paper, we proposed an aspect-based sentiment classification using machine learning techniques on a customer review dataset. The real-time customer review dataset has been considered for the identification of aspects and later for the classification of sentiment. The various feature extraction and selection techniques have been carried out for unique module building, while numerous machine learning (ML) classification algorithms have been used for the classification of aspects as well as the sentiment. Several machine learning (ML) classification techniques such as artificial neural network (ANN), RF (Random forest), Naïve Bayes (NB) and SVM (support vector machine) have used for both classifications. The three feature extraction techniques have been used during the implementation such as TF-IDF, Bigram as well as NLP features. In an extensive experimental analysis, SVM obtains better results with NLP features over other machine learning classifiers.

Keywords: