spacy-annotator

spaCy annotator – a trainable displaCy ENT server

by @tcrossland

This directory contains a customized version of the displaCy REST microservice which includes an endpoint for training the named entity recognition of a model.

The customized version of the service exposes two endpoints:

• POST /ent/ Performs named entity recognition on the text of the request
• POST /train/ Retrains the named entity recognizer of a model using an annotated text

The server can be integrated with an annotation frontend, for example, the experimental spaCy annotator.

To run the server locally, install the dependencies and execute python app.py.

---

POST /ent/

Example request:

{
    "model": "en",
    "paragraphs": [
        {
            "text": "When Sebastian Thrun started working on self-driving cars at Google in 2007, few people outside of the company took him seriously."
        }
    ]
}

Name	Type	Description
model	string	identifier string for a model installed on the server
paragraphs	array	paragraphs to be parsed

Example response:

[
    {
        "tags": [
            { "end": 20, "start": 5,  "type": "PERSON" },
            { "end": 67, "start": 61, "type": "ORG" },
            { "end": 75, "start": 71, "type": "DATE" }
        ],
        "text": "When Sebastian Thrun started working on self-driving cars at Google in 2007, few people outside of the company took him seriously."
    }
]

The response contains an array of paragraphs, each paragraph has the following properties:

Name	Type	Description
tags	array	array of tags (see below)
text	string	the text of the paragraph

Each tag has the following properties:

Name	Type	Description
end	integer	character offset the entity ends after
start	integer	character offset the entity starts on
type	string	entity type

---

POST /train

Example request:

{
  "model": "en",
  "paragraphs": [
    {
      "text": "Banco Bilbao Vizcaya Argentaria (BBVA) es una entidad bancaria española, presidida por Francisco González Rodríguez.",
      "tags": [
        {
          "start": 0,
          "len": 31,
          "type": "ORG"
        },
        {
          "start": 33,
          "len": 4,
          "type": "ORG"
        },
        {
          "start": 63,
          "len": 8,
          "type": "NORP"
        },
        {
          "start": 87,
          "len": 28,
          "type": "PERSON"
        }
        ]
    }
  ]
}

Name	Type	Description
model	string	identifier string for a model installed on the server
paragraphs	array	paragraphs to be parsed

Each paragraph has the following properties:

Name	Type	Description
tags	array	array of tags (see below)
text	string	the text of the paragraph

Example response:

{
  "paragraphs": [
    {
      "tags": [
        {
          "end": 31,
          "start": 0,
          "type": "ORG"
        },
        {
          "end": 37,
          "start": 33,
          "type": "ORG"
        },
        {
          "end": 71,
          "start": 63,
          "type": "NORP"
        },
        {
          "end": 115,
          "start": 87,
          "type": "PERSON"
        }
      ],
      "text": "Banco Bilbao Vizcaya Argentaria (BBVA) es una entidad bancaria española, presidida por Francisco González Rodríguez."
    }
  ]
}

Name	Type	Description
paragraphs	array	parse result for each paragraph

The response contains an array of paragraphs, each paragraph has the following properties:

Name	Type	Description
tags	array	array of tags (see below)
text	string	the text of the paragraph

Each tag has the following properties:

Name	Type	Description
end	integer	character offset the entity ends after
start	integer	character offset the entity starts on
type	string	entity type