Sentimentality

Helps reviews be heard!

Comment

Inspiration

We help businesses use Machine Learning to take control of their brand by giving them instant access to sentiment analysis for reviews.

What it does

Reviews are better when they are heard. We scrape data from Yelp, run the reviews through our ML model and allow users to find and access these processed reviews in a user-friendly way.

How we built it

For the back-end, we used Flask, Celery worker, and Dockers, and TensorFlow for our machine learning model. For the front-end, we used React, bootstrap and CSS. We scraped the yelp data and populated it to a local MongoDB server. We perform periodic Celery tasks to process the scraped data in the background and save the sentiment analysis in the database. Our TensorFlow model is deployed on the GCP AI Platform and our backend uses the specified version.

Challenges we ran into

  • Learning new technologies on the fly during the day of the hackathon. Also, commutation barriers and deployment for machine learning model
  • Training, building and deploying a machine learning model in a short time
  • Scraping reviews in mass amounts and loading them to the db
  • Frontend took a while to make

Accomplishments that we're proud of

  • To get a working prototype of our product and learn a few things along the way
  • Deploy a machine learning model to GCP and use it
  • Set up async workers in the backend
  • Perform sentiment analysis for over 8.6 million reviews for almost 160,000 businesses

What we learned

  • Deploy ML models
  • Performing async tasks on the backend side

What's next for Sentimentality

  • Provide helpful feedback and insights for businesses (actionable recommendations!).
  • Perform more in-depth and complex sentiment analysis, and the ability to recognize competitors.
  • Allow users to mark wrong sentiments (and correct them). Our models aren't perfect, we have room to grow too!
  • Scrape more platforms (Twitter, Instagram, and other sources, etc.)
  • Allow users to write a review and receive sentimental analysis from our machine learning model as feedback
  • Allow filtering businesses by location and/or city
Share this project:

Updates