Continuous Learning with Orby
Orby is engineered to refine its capabilities over time, adapting and improving as it learns from the human-reviewed documents you provide. However, there are key aspects to be aware of to ensure that you are harnessing Orby's full potential: understanding workflow-isolated learning and initial confidence scores.
Workflow isolation : Orby learns separately within each workflow. Imagine you have one workflow for classifying documents and two separate workflows for extracting data based on those classifications. In this case, Orby needs to accumulate multiple human-reviewed examples for each classification type to become proficient at data extraction for both classes.
Confidence scores : Initially, Orby might display low-confidence scores when making predictions. Don't be alarmed! This is normal for a system that learns continuously. As you feed it more examples, it will grow more confident and accurate in its predictions.