BERT Product Rating Predictor

Overview

• Year: 2020
• Language(s): Python, R
• Discipline(s): Machine Learning, Natural Language Processing (NLP)
• Keywords: Amazon Reviews, BERT, Classification, Clustering, Machine Learning, NLP

Description

The BERT Product Rating Predictor is a natural language processing model based on the Bidirectional Encoder Representations from Transformers (BERT) model developed to predict star ratings for textual product reviews. It was created by fine-tuning the BERT model, training it with a custom dataset containing 195,765 reviews gathered from Amazon’s electronic products section. From this model, a k-means clustering model was then employed to explore interesting relationships between the predicted star ratings and their associated reviews.

This model was then used as part of a research project titled Using the BERT Model to Predict and Analyze Star Ratings from Reviews in Amazon Electronic Products.

Team Members

• Alexander Roustai
• Jin Koay
• David Wecke
• Ankita Tripathi
• Carlos Santiago Bañón

Build Instructions

1. Download the bert-product-rating-predictor.ipynb Jupyter notebook.
2. Download the pretrained model.
3. In the notebook, set up the pretrained model by adding the pretrained model to the same directory and including it as the file_path. More instructions can be found in the notebook.
4. Reach the end of the notebook and define a custom review.
5. Run the notebook. The last block will contain your predicted star rating.