An empirical study on text based sentimental analysis of using various machine learning classifiers on newspaper
Harjender Singh Harjender Singh, Hitendera Garg, Ajay Kumar Phogat · Computational Methods in Science and Technology · 2024
Sentiment classification is a way to evaluate or analyse the important subjective information from the text. Analyzing the sentiments in newspaper is crucial for finding the pulse of public opinion i.e. parsing sentiments within newspaper is a complex task. We prioritize models that not only make accurate predictions but also provide clear explanations for their decisions. By comparing different models, we identify approaches that offer both accuracy and transparency and giving more insightful analyses of newspaper content. Our study involves a diverse collection of newspaper articles, and through rigorous evaluation, we highlight models that provide a balance between accuracy and interpretability. Text analysis methods for sentiment analysis typically work at a specific level, such as the pharse level, sentence or document level. This research paper aims to explore the application of TfidfVectorizer and counter vectorization techniques including Random Forest Classifier, Logistic Regression and Naïve Bayes methods, these techniques are evaluated on text based sentence-level sentiment analysis on newspaper headlines. In this paper the dataset used by me the target label is not provided, which should indicate whether the headlines are positive, negative, or neutral. To create the specified column, we can calculate the positive, negative, neutral, and compound scores for each headline. Based on the compound score, we will generate a ‘Sentiment’ column, which will serve as our target column during model training.