This model was trained on english videos ONLY. Due to our imbalanced dataset ("good"
videos are more prevalent than "bad" or "neutral" videos), we focused on various F1 scores,
Accuracy, and MCC (Matthews correlation coefficient) to track model performance.
While we do believe our model will still perform fairly well due to the impact of other features, we can't guarantee any
accuracy metrics on non-english videos, especially due to current limitations in the sentiment
analysis library used. We recognize these elements may introduce bias into the model, and hopefully
can improve it in the future!