Comparative Product Reviews on Table Tennis Rubbers

Purpose

• Our team implements natural language processing models to mine comparative opinions on table tennis products

Tech Stack

• MongoDB, Sqlite3, Python, Flask, Dash, Plotly, Fastapi, CSS, Bootstrap, Heroku

Live Deployment

• On Heroku: http://tabletennis-equipment-showdown.herokuapp.com/

Directory Structure:

• |-- EDA: Folders contain data analysis jupyter notebook
• |-- scraper: webscraping/data ETL module
• |-- comparison: Roberta classification model training using labelled data and inferencing on unlabelled data
• |-- information_retrieval: Extracting Entity, Attribute, Directions from comparative sentences
• |-- common: common utils (connect to db, data query)
• |-- viz: wordcloud visualization
• |-- frontend: dashboard

Create Local Testing Environment

• conda create -n ds4a-t15 python=3.7 pip
• conda activate ds4a-t15
• pip install -r requirements.txt

Fire up Local Front End Environment

• cd frontend/plotly
• python app.py
• Open up Google Chrome Browse to http://127.0.0.1:8888

Fire up local sqlite db

• cd frontend/plotly/db
• sqlite3
• sqlite> .open tableTennisData.db

Contributors

• Charlotte Giang
• Eyan Yeung
• Pooja Umathe
• Jessie Zhang
• Winnie Yeung (Team Lead)

Further Reading

• Slide Deck Presentation for DS4A conference (10/16/2020): https://bit.ly/35eXcqD