strengths
- the model is trained with almost 15.000 comments and the distribution of the rates are balanced
- it can predict the rating more accurate if the actual rating is in between 2 and 4
- capital letters does not affect the output rating

weaknesses
- the model is having trouble with predicting if the given rating is not consistent with comment itself in terms of sentiment
- longer comments tends to be considered as average rating since it increases the neutral score
- typo in the comment is affecting the result
- exclamation mark and emojis are also affects the output rating