Resources

Newsroom

Back to Industry Trends & Insights

Community Model Gathers Document Knowledge to Create Better AI for Everyone

Social Cdm Image

Community = The Power of Many. As customers feed documents into the platform, it gains more knowledge and becomes a smarter, more effective tool for everyone.

By Ken Revett, Principal Data Scientist, AI Foundry

We often tout the advanced AI and ML technology that are part of AI Foundry’s Cognitive Business Automation Platform, but central to this, and perhaps most importantly, is the Community Document Model. What does this community model do? It centralizes “learnings” from the vast documents and data processed by the Agile Mortgages solution. It makes our platform the best that it can be by helping it to gain knowledge and “learn” different to recognize different documents and fields and then share with all users.

Community = The Power of Many

We have enabled the industry to provide the data to digitally transform all documents using our computer vision and deep learning technology. As more customers use our Agile Mortgages solution, and feed more (and different) documents into the platform, they are learned, and that knowledge can be used again in the future. As this community of customers keeps feeding documents into the platform, it gains more knowledge and becomes a smarter, more effective tool for everyone.

At the heart of how the Community Document Model learns, is the ability to sort known and unknown documents. The process of learning and classifying new documents is automated, which provides a crucially important pathway for self-improvement. Our Agile Mortgages Solution will take the unknown documents and group them based on similar features. Our community of customers also play a critical role in helping to annotate the unclassified documents. The combination of these two approaches is used to automatically create groups of data that will be deployed for retraining the Community Document Model.

In order to determine that the newly trained model has learned at a sufficiently accurate level, it is tested (off-line) on documents that were previously unknown. Once that is confirmed, the model is re-deployed, without skipping a beat – ready for the next cycle of deployment-learning-deployment.

Similar to the technology used in the autonomous self-driving car efforts, the big challenge is how to efficiently manage all of the data that is being fed into the model and teaching it how to make decisions faster, under varied conditions.

Since the start of 2020, ten mortgage companies and banks have become members of our community, and more are being added all the time. We expect the community to grow to more than 100 contributors by 2024. Eventually, we will get to the point that over 10,000 different document types and variants will be learned by our system. Think of what a powerful tool this platform can become over time!

If you are considering using AI and ML to enhance and improve your back-office operations, we encourage you to join our community. You will not only benefit from improved productivity, increased efficiency and reduced costs, but will also become a part of building a powerful, growing system that will benefit the entire mortgage industry.

Contact us for more information on our community-based approach.

Want to learn more?
Download the Community Document Model feature brief from our resource library. 

Download brief+