In this project, I classified positive and negative Italian words through Italian articles by using the Least Squares Method, which is also known as Linear Regression. It is important to note that I possess no prior knowledge of the Italian language, ensuring that the project remains as unbiased as possible. In addition to Linear Regression, I also made a logistic regression model to compare the ranking of the semantics, along with a Linear Regression model done without term frequency-inverse document frequency (TFIDF) filtering.
Conclusion: The Linear Regression model managed to capture the sentiments of the words relatively accurately. Although some neutral words were included in both the positive and negative classifiers, a manual inspection of the scores in the dataset reveals that the predicted words consistently align with the average scores among articles containing those words.