Augment Dataset

Augment a dataset by removing duplicates/outliers and generating synthetic samples.

This endpoint creates a NEW dataset with the augmentations applied. Operations include:

remove_duplicates: Remove exact duplicate samples.
remove_outliers: Remove samples with anomalous lengths.
balance: Generate synthetic samples for underrepresented classes/entities.

The original dataset is preserved.

Request body

task_type'ner' | 'classification' | 'custom' required

Task type of the dataset

dataset_namestring nullable

Name of stored dataset to augment

dataset_versionstring nullable

Dataset version (latest if omitted)

datasetobject[] nullable

Inline dataset to augment (if dataset_name not provided)

target_distributionobject nullable

Target class/entity distribution as percentages (must sum to 1.0)

domain_descriptionstring nullable

Domain description for synthetic sample generation

new_dataset_namestring nullable

Name for the new augmented dataset (required if update_in_place=False)

labelsstring[] nullable

Labels for generation (required if balance operation used)

update_in_placeboolean

If True, creates new version with same name and soft-deletes old version

Example request

{
  "dataset_name": "my-reviews-dataset",
  "domain_description": "Customer reviews for e-commerce products",
  "labels": [
    "positive",
    "negative"
  ],
  "new_dataset_name": "balanced_reviews_v2",
  "operations": [
    {
      "enabled": true,
      "type": "remove_duplicates"
    },
    {
      "enabled": true,
      "type": "remove_outliers"
    },
    {
      "enabled": true,
      "type": "balance"
    }
  ],
  "target_distribution": {
    "negative": 0.5,
    "positive": 0.5
  },
  "task_type": "classification",
  "update_in_place": false
}

Response

Successful Response

successboolean required

original_dataset_namestring nullable

original_dataset_versionstring nullable

messagestring required

updated_in_placeboolean

True if the original dataset was replaced (soft-deleted)