Oracle Cloud Infrastructure Documentation
Try Free Tier

Custom Model Details

Document Understanding custom models have metadata and metrics information that you can reference.

Model Metadata Overview

The following metadata is provided for the custom model in Document Understanding.

Custom Model Metadata
Attribute Details
labels

A list of the user-defined labels in the custom model. For document classification, these are the names of the document classes. For key value extraction, these are the names of the custom fields

For example, for the custom key value extraction model for a Bill of Landing document, the labels might be CarrierName, ShipmentID, Recipient, and TrailerNo.
compartmentId The compartment identifier.
definedTags Defined tags for this resource. Each key is predefined and scoped to a namespace. For example: 
{"foo-namespace": {"bar-key": "value"}}
description (Optional) A description of the model.
displayName A human-friendly name for the model, which can be changed.
freeformTags A simple key-value pair that's applied without any predefined name, type, or scope. It exists for cross-compatibility only. For example: 
{"bar-key": "value"}
id A unique identifier that's immutable after creation.
lifecycleDetails A message describing the current state in more detail. It can provide actionable information if the training failed.
lifecycleState

The current state of the model.

  • CREATING
  • UPDATING
  • ACTIVE
  • DELETING
  • DELETED
  • FAILED
modelType

The type of Document model this is.

Allowed values are:

  • DOCUMENT_CLASSIFICATION
  • KEY_VALUE_EXTRACTION
projectId The OCID of the project that contains the model.
systemTags Usage of system tag keys. These predefined keys are scoped to namespaces. For example: 
{"orcl-cloud": {"free-tier-retained": "true"}}
timeCreated When the model was created, as an RFC3339 datetime string.
timeUpdated When the model was updated, as an RFC3339 datetime string.

Model Metrics Overview

The following metrics are provided for custom models in Document Understanding

Model Metrics
Attribute Details
Precision The fraction of relevant instances among the retrieved instances.
Accuracy The fraction of correct predictions among the total predictions.
Recall The fraction of relevant instances that were retrieved.
Threshold The decision threshold to make a class prediction for the metrics.
Total documents The total number of documents used for training and testing.
Test documents The number of documents from the dataset that were used for testing and not used for training.
Training duration The length of time in hours that the model was trained.

Was this article helpful?