Data Set

Below is an example from the Amazon Consumer Reviews of Amazon Products data set showing the review rating (1-5) and the correlating review text for a single product in the Amazon data set. Amazon reviews should be at least 20 words and limited to 5,000 words. They are written typically in informal English, contain irregular expressions, and can contain abbreviations and slang words. Amazon sentiment analysis approaches can be generally categorized into two main approaches: machine learning and lexicon-based approach.

Data Distribution

Data set used in this study has a total of 34,626 records.

Data is imbalanced as most reviews are of class 4 or 5. This issue was overcome by over sampling the data with lower class labels.

Over Sampling of the Data

With the unbalanced data set, the second model is trained by oversampling class labels with lower count.

Cleaning of Data

• Clean data by removing special characters as it does not offer any meaningful context to the review.
• Replace contractions with it’s meaningful counterpart.
• Natural Language Processing also takes advantage of stop words ( common words which would appear to be of little value ) and stemming(replacing words with its root words).
• For sentiment analysis purposes we will not be removing stop words since it removes word like ‘not’ and can potentially change the meaning of the review.

Visualize Data

Certain words are used more often than others.

Split Data for Machine Learning

Split data into a training, test, and validation set.

This is a list of over 34,000 consumer reviews for Amazon products like the Kindle, Fire TV Stick, and more provided by Datafiniti's Product Database. The dataset includes basic product information, rating, review text, and more for each product.