Bringing the Human-in-the-loop needed for AI with Figure Eight

Machine learning requires labels, or annotations, to tell algorithms what the right answers are in the training phases of model development. But it’s difficult to build a fully annotated data set, because you can’t just rely on AI tools to annotate your training data, you also need humans to label data.

To advance our data annotation capabilities, we are excited to share that IBM Watson Studio is partnering with Figure Eight, a platform that leverages “human-in-the-loop” data labels to train and tune models. Figure Eight’s industry leading platform transforms unstructured data (text, image, audio, video) into customized training data.

Ruchir Puri — IBM Fellow and Chief Architect, IBM Watson — on how to Reinvent Businesses with AI

IBM Watson Studio is directly integrating Figure Eight’s templates for image categorization, natural language classification, sentiment analysis, and content moderation, making it seamless for you to get high quality data to train and tune your models. The integration into IBM Watson Studio will enable a feedback loop, so new data can be continuously annotated and leveraged, thus improving data drift, model optimization, and model performance. You can see the integration in action here:

Demo of Figure Eight integration in IBM Watson Studio

IBM Watson Studio customers now have the ability to sync their data to Figure Eight via an API and will have full access to the features and functionality of the Figure Eight platform. This includes the ability to code custom logic, write unique instructions for workflows, enjoy the full suite of quality controls, access their pool of skilled annotators, and more. Once data is annotated, users can seamlessly connect their data back into IBM Watson Studio and train model with the multiple tools available.

Learn More