tag_svc_model.md

Tag Service Data Model

The Tag Service persists data in Mongodb. Here we list information about the data model.

Collections

The Tag Service supports several classes of data, each with its own collection in Mongo:

• model - stores information about a Machine Learning model that is known to the system.
• tagging_event - represents a specific event where tagging of assets occured. An model can have zero or more tagging_events.
• asset_tags - represents all tags for a given asset, storing tag text, confidence and tagging event. The asset_uid relates to any type of asset, but we intend for this to commonly be a bluesky datum object uid.

Entity Relation Diagram

The following ER Diagram is slighltly modified from traditional ER diagrams to accomodate a document database like Mongo. In some cases, the model uses PK/FK-style relationship, in others it uses sub-documents to establish relationships.

Example Documents

asset_tags = {
    "asset_uid": "ee600210432b8f81ad229c33",
    "sample_id": "house paint 1234",
    "tags": [
        {
            "tag": "rods",
            "confidence": 0.9008,
            "event_id": "wwewere6002104rwerwe81ad229c33",
        },
        {
            "tag": "peaks",
            "confidence": 0.001, 
            "event_id": "wwewere6002104rwerwe81ad229c33",
        }
    ],
}

tagging_event = {
    "uid": "wwewere6002104rwerwe81ad229c33",
    "tagger_id": "dsfosf9080s9uflkwsnerl3k4j",
    "run_time": 1134433.223,
    "accuracy": 0.7776
}

tagger = {
    "type: "model | humam",
    "model_name": "PyTestNet",
    "uid": "dsfosf9080s9uflkwsnerl3k4j",
    "create_time": 11112333.3  
}