Stanza NLP Service

Features

• On-Demand Model Downloads: Automatically downloads necessary Stanza models for different languages as required.
• Pipeline Caching: Quick second-response times by caching pipelines.

Usage

Run the Docker Container

docker run -p 5000:5000 -v stanza_resources:/root/stanza_resources vivalence/dockerized-stanza-nlp

Run Docker Compose

version: '3'
services:
  stanza-nlp:
    image: vivalence/dockerized-stanza-nlp
    ports:
      - "5000:5000"
    volumes:
      - stanza_resources:/root/stanza_resources
volumes:
  stanza_resources: # model cache dir

Data Structure / API Interface

• POST request to /nlp endpoint with the following JSON structure:

  {
    "language": "en",
    "text": "Hello World",
    "processors": "tokenize,mwt,pos,lemma,depparse"
  }
      

Response

• The service returns a JSON response containing processed NLP data:

  {
    "sentences": [
      {
        "text": "Hello World",
        "tokens": [
          {"text": "Hello", "lemma": "hello", "pos": "INTJ"},
          {"text": "World", "lemma": "world", "pos": "NOUN"}
        ],
        "dependencies": [
          {"dep": "root", "governor": 0, "dependent": 2},
          {"dep": "discourse", "governor": 2, "dependent": 1}
        ]
      }
    ],
    "entities": []
  }
      

  • Includes tokenized sentences, POS tags, lemmas, and dependency parse information.