User Comment Classification Using Linguistic and Sentiment-Based Features

Abstract

The online e-commerce market is growing and becoming increasingly competitive. There are many of the data that businesses provide includes client input, such as product and service reviews. However, customer reviews have a crucial role in the business development and have been valuable sources for marketing intelligence. This paper focuses on explore the effectiveness of using feature extraction during the data preprocessing stage to enhance the performance of learning algorithms. In particular, the experiments were ongoing to investigate the impact of using feature extraction in the classification outcomes with several machine learning models. Three new columns extracted from the features were utilized with five classification algorithms during data preprocessing to classify the sentiment of Amazon reviews. The results are referring to advantages of using the feature extraction which helps making accurate models.  However, Random Forest classifier achieved the best performance among other techniques across both experiments. Addition to that, for Naive Bayes classifier there is no improvement in the model accuracy.