Topic Modeling project

In this project, we are going to train the BERTopic model to identify topics from e-commerce clothing reviews. We are going to use the E-commerce Clothing Reviews dataset, available on Kaggle. It provides real commercial data of an e-commerce website with reviews provided by customers. There are fields, like clothing id, age of the client, title, review text, and so on.

Detailed description of the project

The article with the explanations is Topic Modeling for E-commerce Reviews using BERTopic.

Tools used in the project

• DVC
• DagsHub
• MLflow

Project Structure

• data/: contains all the data
  • raw_data/: contains original data
  • processed_data/: contained processed data
• model/: contains artifact of BERTopic model
• output/: contains the plots generated with BERTopic model
• src: contains the following scripts
  • train.py: Python script to train BERTopic model and save artifact
  • mlflow_log.py: Python script to track the experiments of the ML model
  • topic_model.py: Python script to create BERTopic model
  • process_data.py: Python script to clean and filter the data

Visualize Topics

topic_model.visualize_topics()

topic_model.visualize_barchart(top_n_topics = 10)

topic_model.visualize_documents(docs)

topic_model.visualize_hierarchy()

topic_model.visualize_heatmap(n_clusters=10, width=1000, height=1000)