The quality of both training and evaluation datasets prepared for a specific AI-powered application has significant implications for the quality and Explainability of the AI (Artificial Intelligence) models in production, at an enterprise scale. Enterprise AI practitioners (including researchers, data scientists, developers, engineers, and analysts) struggle with “crossing the valley of death”1, as well as with model and dataset analysis which are often difficult, unsystematic, and unchartered – even the most experienced practitioners need help.
Introducing Azimuth, from ServiceNow.
Azimuth is an open-source software application that helps AI practitioners better understand their dataset and model predictions by performing thorough dataset and error analyses. The application leverages different tools, including robustness tests, semantic similarity analysis, and saliency maps, unified by concepts such as smart tags and proposed actions. Check out the User Guide for a detailed look at all the features and capabilities.
The latest version 2.1.1 of Azimuth is focused on text classification (Natural Language Processing) problems and can easily be adapted to suit other data types and models such as vision or tabular use cases. The Azimuth project’s backlog is open for review, and Community contributions are welcome. Visit the discussion board if you have any questions about using the software or would like to get involved in the project.
Getting started is as easy as following steps 1-2-3. Simply install the latest release of Azimuth with Docker, learn the basics, and try out the tools on your own data. Watch the Getting Started video now.
Azimuth started as a grassroots initiative by a few applied research scientists at Element AI, emerged as a powerful set of tools for AI practitioners through a company hackathon after the acquisition by ServiceNow2, and was then released to the open-source software community under the Apache-2.0 license. If you would like to use Azimuth for your own projects, please be so kind as to cite it accordingly.
“I have been dreaming of building something like Azimuth for a long time after spending (painful) years in Jupyter Notebooks re-writing the same functions and visualizations. I am so excited we can now share those ideas with the Open-Source community!” Gabrielle Gauthier Melançon, Applied Research Scientist and co-founding contributor to the project.
This article was contributed by Sean Hughes and Di Le, with special thanks to Nikola Simic for creating the Azimuth promo video, Frédéric Branchaud-Charron for the getting started walkthrough video, and to our content reviewers and collaborators Gabrielle Gauthier Melançon, Michael Lanoie, and Sethu Meiyappan. This project would not have been possible without the dedication and contributions of the Azimuth team.
© 2022 ServiceNow, Inc. All rights reserved. ServiceNow, the ServiceNow logo, Now, and other ServiceNow marks are trademarks and/or registered trademarks of ServiceNow, Inc. in the United States and/or other countries. Other company names, product names, and logos may be trademarks of the respective companies with which they are associated.