Identify Biasness In Sentiment Analysis
Maria Nusrat, Muhammad Asfand-e-yar · 2024
The corporate environment has gotten extremely competitive in recent years. Understanding people’s feelings and opinions is important. This is where sentiment analysis comes in. Sentiment analysis can be skewed by biases, especially when it comes to things like gender, race and age. Our study explores these biases and checks out how much they affect the results we get from sentiment analysis. To analyze the impact of biases in sentiment analysis, around 10, 000 sentences were compiled, each reflecting specific biases related to gender, age and race. This extensive dataset was then used to evaluate how different models like XGBoost, Naive Bayes, Logistic Regression, particularly Support Vector Machines (SVM) perform in sentiment analysis under the influence of these biases. The goal was to really understand how these biases affect the accuracy and reliability of sentiment analysis. SVM with RBF Kernel was incredibly accurate, scoring 97% in multiclass data and an even more impressive 99% in binary data. SVM with RBF kernel is good, especially considering the dataset, which was full of biases. When it came to a biased dataset, SVMs performed considerably better than the other models.