MLdata-Basic-Schema.json

{
    "$schema": "https://json-schema.org/draft/2019-09/schema",
    "$id": "",
    "title": "ML Data Basic Schema",
    "description": "Information about a set of scientific training data for applying it to machine-learning. This schema servers as a basic for structuring this information and can be extended.",
    "required": [],
    "definitions": {
        "checksum": {
            "type": "string"
        },
        "termDefinition": {
            "description": "Provides information about the human-readable term for a feature or a label, which is optimally defined in a vocabulary. This would support an automated re-featuring or -labeling in case the data needs to be harmonized with other training data. Additional properties could be implemneted to specify for example pre-prpcessing steps for the features and labels.",
            "type": "object",
            "required": ["term", "termValueDataType"],
            "properties": {
                "term": {
                    "type": "string"
                },
                "termDefinition": {
                    "type": "string",
                    "format": "uri"
                },
                "termValueDataType": {
                    "description": "Defines the data type of the feature or label value, on which the data preprocessin steps depend.",
                    "type": [
                        "integer",
                        "number",
                        "string"
                    ]
                }
            },
            "additionalProperties": true
        },
        "annotation": {
            "type": "object",
            "properties": {
                "annotationFileRef": {
                    "type": "string",
                    "format": "uri"
                },
                "annotationChecksum": {
                    "$ref": "#/definitions/checksum"
                }
            }
        },
        "tag": {
            "type": "string"
        }
    },
    "type": "object",
    "required": ["dataLocation", "dataFormat", "dataChecksum", "dataLicense"],
    "properties": {
        "dataLocation": {
            "type": "string",
            "format": "uri"
        },
        "dataFormat": {
            "type": "string"
        },
        "dataChecksum": {
            "$ref": "#/definitions/checksum"
        },
        "dataLicense": {
            "type": "string"
        }
    },
    "featureProperties": {
        "description": "The input features of the data for the machine-learning algorithm. Properties will depend on the particular data.",
        "type": "array",
        "items": {
            "type": "object",
            "required": ["featureTerm"],
            "properties": {
                "featureTerm": {
                    "$ref": "#/definitions/termDefinition"
                }
            },
            "additionalProperties": true
        }
    },
    "labelProperties": {
        "description": "The features that represent the output variables (i.e. labels) and will be predicted by the input features.",
        "type": "array",
        "items": {
            "type": "object",
            "required": ["labelDescription", "labelTerm"],
            "properties": {
                "labelDescription": {
                    "type": "string"
                },
                "labelTerm": {
                    "$ref": "#/definitions/termDefinition"
                },
                "imageLabelType": {
                    "description": "This is an example for image data, how the label information can be more specified. Additional properties could be implemented for other data types, e.g. text data.",
                    "type": ["string", "object"],
                    "enum": [
                      {"$ref": "#/definitions/annotation"},
                      {"$ref": "#/definitions/tag"}
                    ]
                },
                "additionalProperties": true
            }
        }
    }
}