The Implementation of a Hybrid Sentiment Classification for Yoruba Book Reviews

You can download this material now from our portal

The Implementation of a Hybrid Sentiment Classification for Yoruba Book Reviews


The emergence of Web technology resulted in a vast amount of raw data, as users began posting opinions, reviews, and comments online. Extracting valuable information from this data can be challenging. Sentiment Analysis involves understanding and classifying the emotions and opinions expressed by users. The study focused on opinion mining as a text classification task using unigrams as a feature set. The research conducted several experiments, which can be grouped into three categories. In the first group, a lexical classifier algorithm was developed to classify reviews based on the count of opinion words. The algorithm’s performance was evaluated by comparing its results with the actual labels of the reviews. In the second group of experiments, three popular feature selection methods (Chi-Square, Mutual Information Gain, and Galavvotti-Sebastiani-Simi coefficient) were compared for their ability to select a better subset of features. These feature selection methods were combined with three supervised classifiers (Naive Bayes, Logistic Regression, and Support Vector Machine) using different numbers of features (750, 1000, 1250, and 1500). The goal was to find the best combinations of feature selection methods, classifiers, and the number of features for the domain.

In the third group of experiments, the researchers combined the lexical classifier with machine learning techniques sequentially. They performed hybrid sentiment classification to classify Yoruba book reviews as positive or negative. The dataset consisted of 600 Yoruba book reviews collected from various sources like Facebook, personal blogs, and individual book readers.

The results of the machine learning experiments showed that the Naive Bayes algorithm, using the Mutual Information Gain feature selection method and 1500 features, performed the best with an accuracy of 93.33%. The hybrid approach achieved an accuracy of 87%, outperforming the lexical approach (74%), but it did not match the accuracy of the machine learning approach.

The Implementation of a Hybrid Sentiment Classification for Yoruba Book Reviews, GET MORE COMPUTER SCIENCE PROJECT TOPICS AND MATERIALS

Leave a Reply