Background image for Customer feedback automation with Natural Language Processing (NLP)

  • AUTOMATION
  • /
  • NLP

Customer feedback automation with Natural Language Processing (NLP)

  • Zappi are a marketing research company that had been working on a ML model to classify free text feedback on and off for 2 years
  • They found Kortical and decided to test the platform against that tricky NLP problem
  • Using the Kortical AutoML platform they got a 43.9% accuracy gain in a few hours
  • The full ML solution was deployed live in 5 weeks
1

Background

Zappi is revolutionising the market-research industry through automating much of the survey process, enabling their clients to get a more comprehensive service, faster and at a lower cost.

An area they had been spending a lot of time working on was the free text fields in the marketing survey form and they were exploring how to switch from a manual process to a Machine Learning approach.

2

Objectives

The Zappi data scientists wanted a Natural Language Processing (NLP) model that would not only beat their current model but also beat the manual outsourcing team.

3

Kortical AI as a Service Platform

The training data was ready to go as they had been working on this problem outside of the platform. Each text entry had 180 different labels which could apply so it was a complex problem vs a simple positive or negative classifier you hear of often which is not nuanced enough for Zappi’s requirements.

img-cs-zappi-timeline/img-cs-zappi-timeline.png
4

The Results

Using the Kortical AutoML platform they got a 0.59 vs 0.41 - a 43.9% accuracy gain in a few hours

5

Production-ready Machine Learning

Going from ML model to production-ready, is typically a long process (Gartner has reported up to 4 years), as AI models have been built for research and academia vs being able to scale to millions of users and meet software SLAs that organisations need to adhere to. This problem Kortical addresses by providing deployment capabilities as part of the platform. Which meant that the Zappi developers were able to test the best model via API immediately and had the service that the API linked to fully tested and live in production within 5 weeks.

Being an end to end platform made all the difference in the speed that the Zappi data scientists and developers were able to move at.



If any of these points interest you, do get in touch with us via the form below.

Up next

View our other extended client case studies…

View All
Get in touch background

Get In Touch

Whether you're just starting your AI journey or looking for support in improving your existing delivery capability, please reach out.

By submitting this form, I can confirm I have read and accepted Kortical's privacy policy.