ReviewSnap by Team FintasTech

"Make the correct choice in a snap". An NLP based review summarizer for e-commerce products that analyzes reviews and highlights key metrics: Strengths, Weaknesses, Functionality, Price Worthiness.

Comment

Inspiration

We search for so many products, each got so many reviews, we confused sia.

Today’s consumer has a plethora of product choices thanks to globalization and price convergence. One is always surrounded by products and companies. Online shoppers and consumers often face a time-consuming and tedious task of sifting through numerous product reviews in order to make an informed purchase decision. This process can be overwhelming, as it requires them to carefully read and evaluate a large number of reviews in order to identify the strengths and weaknesses of a product, as well as other key metrics such as worthiness, functionality, and overall satisfaction. As a result, they may be limited in their ability to consider a wide range of products, and they may miss out on important information that could influence their decision. This can cause consumer fatigue and secondary effects might lead the consumer to browse fewer number of products.

What it does

Aims to summarize tens of thousands of reviews and highlights key metrics for each product. Using smart classification of reviews based on stars/rating/helpfulness, we will extract and abstract info about product strengths, weaknesses, functionality etc.

How we built it

  • NLP text-davinci-003 framework
  • Recursive Summarization
  • Review Sampling
  • React/Flask stack.

Challenges we ran into

  • Difference of opinions, execution difficulties, deployment was very hard too!
  • Finding a long text summarization model.
  • Dealing with massive review data sets and making the correct tuning of parameters.
  • Managing length in recursive model.

Accomplishments that we're proud of

Sticking together to come up with this idea, figure out ML framework and design of output metrics.

What we learned

A lot, can't even list here! But here's a few:

  • Plan to Fail, and then plan again until you unexpectedly succeed.
  • Have a vision, and don't compromise on it till the deadline.
  • Stick together and respect each other's work.
  • Document everything and make it usable/accessible to other coders.

What's next for ReviewSnap by Team FintasTech

Become a unicorn by 2024. Nothing less.

Share this project:

Updates