Context This dataset was used for my undergraduate research. The main problem in this dataset is its imbalanced nature. Content I scraped the website of Coursera and pre-labelled the dataset depending on their rating. For a 5-star rating, the review was labelled as Very Positive, Positive for 4-star, Neutral for 3-star, Negative for 2-star, and Very Negative for 1-star. There are 2 files, **reviews.tsv** and **reviews_by_course.tsv**. The **reviews.tsv** file has no grouping, just the course reviews and their corresponding label. For the **reviews_by_course.tsv**, they are grouped by the CourseId column. reviews.tsv **Id** - The unique identifier for a review. **Review** - The actual course review. **Label** - The rating of the course review. reviews_by_course.tsv **CourseId** - The course tag. This is in the URL of the course in the Coursera website. For example, in this URL, [**machine-learning** would be the course tag][1]. **Review** - A review in a specific course. **Label** - The rating of the course review. [1]: https://www.coursera.org/learn/machine-learning