This project is an advanced implementation of a product recommendation system that leverages the power of Sentence Transformers. Initially, the recommendation system was built using KMeans and TF-IDF vectorizer, but in this enhanced version, we aimed to improve the recommendation quality by utilizing the latest natural language processing techniques.
A Sentence Transformer is a type of natural language processing model designed specifically to produce meaningful and useful sentence embeddings. Sentence embeddings are fixed-length numerical representations that capture the semantic meaning of a sentence.
The traditional recommendation system based on KMeans and TF-IDF has limitations in capturing semantic and contextual similarities between products and user preferences. By adopting BERT and Sentence Transformers, we can overcome these limitations and provide more accurate and context-aware product recommendations.
The project was undertaken during an 8-week-long summer workshop held at the Center for Innovation and Entrepreneurship (CIE), PES University RR.
The team behind this innovative project comprises:
- Harish Satheesh
- Kota Bharadwaj
- Sahana
- Vishal
Center for Innovation and Entrepreneurship (CIE): Link to CIE
- Utilizes BERT and Sentence Transformers for semantic understanding and context-aware recommendations.
- Improves recommendation quality over the previous KMeans and TF-IDF based approach.
- Supports custom user preferences to provide personalized recommendations.
- Easy-to-use API for integrating the recommendation system into your applications.
- Extensible and modifiable architecture to accommodate specific use cases.
You can find a detailed step-by-step implementation of the recommendation system in the Jupyter Notebook provided in this repository: Link to Jupyter Notebook
The dataset(CSV file) used for training and evaluation can be downloaded from the following link: Link to Dataset
Clone the repository to your local machine:
git clone git@github.com:VishalS-HK/product-recommendation-system-BERT.git
cd product-recommendation-system-BERTFor any inquiries or questions, please contact vishal.rsn21@gmail.com.
