Govtext

Team - Data science and AI team Govtech
Platform - Web
Role - UX designer with 1 creative designer, 1 product owner, 2 data scientist, 2 software engineers
Duration - 12 Months to MVP.

What are we trying to solve?

Public officers have a tough job sorting through tons of articles/feedback to find common themes and make detailed reports. They have limited access to commercial websites due to security and privacy rules.

Usually, they carefully go through documents and put together short presentations. Also, they analyze data and feedback to find recurring themes and long-lasting topics.

What did we do?

By mapping our the journey of our user, we were able to have a holistic view of their pain points and opportunities for us to improve the existing legacy platform.

Solutioning

Building on top of Govtext 1.0 we revamped it to 2.0 a Natural Language Processing-based text analytics platform to help officers, even those without coding experience, analyze text. They can easily look for common topics of interest and find specific documents containing certain words.

We also made it possible for officers to summarize their findings or articles quickly.

Success and outcome

• Helped officers that deal with textual datasets increase efficiency/productivity in generating results based on surveys conducted after release.

• Encouraged and enabled officers with no coding background to analyse large samples of feedback or any textual datasets at ease.

Reflections

This project began as an update of an old platform that had been neglected.
We didn't just want to improve the look, so we looked for new opportunities. There were many discussions about changes to the user interface and the ideas that data scientists were exploring.

Looking back, I'll list some of the challenges we faced.

• How to visualise text-analytics and make sense of it

• Userflows/enhancements were made

• Implementing Hotjar/WOGA (Whole of government analytics) to track issues and user flows

• Collating all possible error/interaction feedback messages from back-end/input to users