The Sentiment Landscape: Final Insights

The elimination of phone chargers from the packaging of new phones has catalyzed a broad discussion of sustainability, corporate accountability, and consumer rights. This project utilized text analysis and machine learning methods to study systematically how various media sources frame this policy change. By applying various supervised learning models to a dataset of news articles, sentiment classification for charger removal was possible as Pro, Neutral, or Against, showing trends in broader public discussion. The findings offer valuable insight into the way that modern media covers environmental claims made by tech firms.

​

The study started with classical approaches such as Naïve Bayes, which was used as a baseline model to set initial performance metrics. Naïve Bayes, despite its simplicity, performed reasonably well but had difficulty in distinguishing between more subtle supportive or critical sentiment, although it was good at recognizing neutral sentiment. Decision Trees, when used, added insight through the provision of explicit and interpretable visual representations of decision pathways based on word attributes. These initial models illustrated that even relatively simple machine learning techniques can introduce useful structure into apparently subjective material such as media sentiment.

​

In order to surpass limitations of more basic models, more advanced approaches like Support Vector Machines (SVMs) and Neural Networks were employed. The SVM with a Radial Basis Function (RBF) kernel achieved the highest accuracy, demonstrating its ability to recognize complex boundaries between intersecting sentiment classes. Likewise, Neural Networks created with Keras showed high effectiveness, successfully modeling non-linear relationships in the text data converted with Term Frequency-Inverse Document Frequency (TF-IDF). Despite the inherent challenges posed by the size of the dataset for deep learning models, the neural network’s capacity to accurately identify neutral reporting with significant precision confirmed its importance in advanced sentiment analysis frameworks.

​

Throughout this inquiry, the pattern was generally the same: Media reporting was mostly neutral in tone, with critical or supportive voices fewer but not absent. This pattern may be observed to reflect journalistic principles of objectivity or corporate communicative strategy planned to positively spin environmental action. Regardless of origin, machine learning algorithms picked up on these patterns more clearly and efficiently than might be possible using manual screening. Visualization tools such as confusion matrices and decision trees gave concrete means of interpreting the output of the models, hence boosting confidence in the output.

​

​In conclusion, this study highlights the promise of modern data science techniques for enhancing understanding of complex social discourses. The combination of traditional statistical frameworks with sophisticated neural networks enables the uncovering of hidden patterns in large datasets of text data. Such instruments are becoming ever more critical to the careful assessment of corporate sustainability claims, public policy debates, and media framing strategies. As machine learning capabilities continue to improve, projects like this demonstrate the immense potential for using automated analysis to foster transparency, accountability, and data-driven decision-making in a wide range of industries and issues.