spaCy models for NER

This ML backend provides a simple way to use spaCy models for Named Entity Recognition (NER) and Part-of-Speech (POS) tagging.

Current implementation includes the following models:

• Named Entity Recognition (NER)
• [coming soon…] Part-of-Speech (POS) tagging

Quickstart

1. Build and start the ML backend on http://localhost:9090

docker-compose up

2. Validate that the backend is running

$ curl http://localhost:9090/health
{"status":"UP"}

3. Create a project in Label Studio. Then from the Model page in the project settings, connect the model. The default URL is http://localhost:9090.

Usage

Labeling configuration

This model is compatible with the following labeling configurations:

<View>
    <Labels name="label" toName="text">
        <Label value="CARDINAL" background="#FFA39E"/>
        <Label value="DATE" background="#D4380D"/>
        <Label value="EVENT" background="#FFC069"/>
        <Label value="FAC" background="#AD8B00"/>
        <Label value="GPE" background="#D3F261"/>
        <Label value="LANGUAGE" background="#389E0D"/>
        <Label value="LAW" background="#5CDBD3"/>
        <Label value="LOC" background="#096DD9"/>
        <Label value="ORG" background="#ADC6FF"/>
        <Label value="PERSON" background="#9254DE"/>
        <Label value="TIME" background="#F759AB"/>
    </Labels>
    <Text name="text" value="$text"/>
</View>

You can also use the default configuration from the Named Entity Recognition template provided with Label Studio.

Note: If your labels are different from the default ones, the text spans will still be highlighted, but you have to manually map the labels to the ones you have in the model. Go to model.py and change _custom_labels_mapping to map from SpaCy entities to your labels. Check for predefined labels in the official SpaCy documentation, for example for en_core_web_sm.

Parameters

To change default parameters, specify the following environment variables:

• PORT - port to run the server on, default is 9090
• WORKERS - number of workers to run the server with, default is 2
• SPACY_MODEL - spaCy model to use, default is en_core_web_sm