Thomson Reuters Deep Learning Technology Model

Project Type: UX Design, Service Design

Design Problem: Thomson Reuters has a patent for their Text Mining Model which uses machine learning to scan through data from news articles and identify the level of risk that a specific company might default, rendering a “default score” that investment professionals use in their financial analysis. While the original model was based on individual key words, the default score proved to be unreliable, so TR improved the model with a new technology called deep learning, which interprets whole phrases or sentences of an article to increase the default score accuracy. The challenge was to build a new UI and experience demonstrating the value of the new model within Eikon, TR’s financial product.

Process: Personas, Journey Maps, Sketches, Low Fidelity, High Fidelity, Prototype, User Research. Based on internal matter experts’ knowledge of the users, I defined initial personas and journey maps. Using the existing product UI standards, I sketched out 2 UI solutions which was tested internally first, and iterated. I prepared a research study to interview 5 current users of the Text Mining Model.

Final Solution: A new UI and experience within the Eikon platform that displays the new Deep Credit Model and emphasizes on the new default score robustness.

Limitations: Budget for User Research.

Applications: Sketch, Invision, Keynote

Team: Worked with the Head of Data Science and with a Design Strategist.

Sketches