UX and UI Design
Appen Limited
Appen Limited is a publicly traded data company listed on the Australian Securities Exchange under the code APX. Appen provides or improves data used for the development of machine learning and artificial intelligence products. I was responsible for creating interfaces for Natural Language Processing annotation tools. NLP helps computers communicate with humans in their own language and scales other language-related tasks.
Work developed at Novatics
Audio Tools
AT is a set of audio tools available at Appen to collect data for AI projects and more. This tool is used by clients such as Facebook and Amazon. Our clients set up jobs for skilled contributors to annotate, transcribe and translate audio excerpts. I worked as the UI/UX designer for the team, dedicated to making our tools even better. This project was highly complex not only because of the nature of these tasks, but also because it is a product with millions in revenue.
Contributors are paid hourly rates based on their efficiency and time spent on each job. So the goal is to allow them to be as efficient as possible. We want contributors to work fast with high quality – and save us money in the process.
A few of my contributions to the tool:
🌍 Individual Waveform
and Time Stamping
In audio transcription mode users had to wait a long time while the audio loaded. And they also couldn't see a more clear and detailed waveform for each segment, so they weren't able to easily identify speech/silent intervals. Both were complains that I identified with my Project Manager help in running user interviews and studies.
To solve this I suggested the Individual Waveform for each segment. This way users would be able to scan faster across the audio excerpt and annotate faster. The tool would also load tiny bits of the audio instead of its entire length.
New solution
Previous solution
The handoff for this project demanded a lot of time because of all the interactions that must be precise. In my handoff files my goal is to make myself as clear as possible so the dev team can work just fine.
Here you can see how the handoff for timestamping was delivered. Step by step described.
Text Tools
This set of tools has the same goal as Audio Tools, but here users go through excerpts of texts annotating what is needed. My biggest contribution for this tool was Side by Side Arbitration.
🌍 Side by Side Arbitration
For the design it was an extensive and complex project. The biggest challenge was how to fit so much information in such a small workspace. And it gets more complex with all the necessary interactions we had to offer.
This is the "Disagreement + Audit" workflow and it consists in allowing the tool to display two annotations that have inconsistencies so a third person can audit to chose the correct one or enter a new version. This mode is very popular amongst other NLP tasks.
Wireframing
It took several iterations to reach a good solution for this project. As usual I started by understanding the problem and gathering as much info as possible. I also got in touch with a few users to hear their POV.
With all the information gathered, I moved on to wireframing and rapid prototyping.
When the team felt confident enough with the concept we worked on, I moved on to the final design.
High fidelity prototype in Figma
