1 of 13

Sift

Danny Nguyen, Isabel Abonitalla, Thuy Le, Taylor Hines (Team Ditto)

Helping restaurant owners sift through the noise and gain constructive feedback.

2 of 13

  • Problem statement
  • User persona
  • Our solution
  • Machine learning architecture
  • Front-end designs
  • Links
  • Thank you

Agenda

3 of 13

The Problem

3

Quantity of reviews

Hard to digest feedback

Difficulty pinpointing customer issues

4 of 13

This is Gunther’s Ice Cream in Sacramento, CA

How might we help small to medium sized businesses, like Gunther’s in the Sacramento area and beyond, efficiently gain insights from their online reviews?

4

5 of 13

For example, Gunther’s Ice Cream has 2,000+ reviews on Yelp

Reading and trying to get helpful feedback from each review would take a long time

5

6 of 13

Our Solution

A web platform powered by machine learning that helps business owners...

6

Sort through thousands of reviews quickly to save time

Visualize data for KPIs

Aggregate high level insights from reviews

7 of 13

Natural Language Processing allows us to extract entities (keywords like the restaurant’s dishes or other features) then perform sentiment analysis to determine whether the entity is spoken of positively or negatively (in this case, within the reviews).

Powered by Machine Learning

We used IBM Cloud’s Natural Language Understanding API

7

8 of 13

Our code

8

9 of 13

Team Ditto was blocked by an API paywall and was unable to build and hook up the following front-end designs to the working machine model, but this is what we hoped to implement.

9

10 of 13

Front-end Designs

Sift offers quick high level visualizations

10

11 of 13

Front-end Designs

At a glance metrics with more information if needed

Generated word cloud from text strings in reviews

11

12 of 13

Links

12

13 of 13

Thank you SacHacks 2021!

13