Model Validation Tasks (MVTs) improve the Document Classification, Field Identification, and Table Identification automation rates. During model training, a model can make high-confidence predictions that are incorrect. To minimize incorrect predictions, you need to complete MVTs. The results from the MVTs can be used for re-training the model to achieve greater automation.
Types of Model Validation Tasks
In v37, you can complete Document Classification MVTs, which prompt you to classify documents manually. To learn more about Document Classification MVTs, see Document Classification Model Validation Tasks.
Previous versions of Hyperscience included Field ID MVTs and Tabe ID MVTs, but these tasks have been replaced by the Anomaly Detection feature in v37. When you analyze the training documents for a Field Identification or Table Identification model, the system identifies anomalies in your annotations of fields and table columns. You can then review and correct the annotations, if necessary. To learn more about Anomaly Detection, see Detecting and Correcting Anomalies in Field Annotations and Detecting and Correcting Anomalies in Table Annotations.
Additional Notes
Document Classification MVTs are identical to QA tasks, so they will appear in the same queue.
When a new Classification model is trained or imported, the old MVTs are deleted.