transner

NER with transformer

route: /transner/v0.3/ner

• input: JSON object containing a list of strings {“strings”: [...]}

  • This interface expects sentences taken eventually from longer documents or records from a table. Please check with Pipple if they are willing to contribute to the provision of a sentence splitter for longer documents. Otherwise, we will implement it ourselves

• output: JSON object containing the extracted entities

• example of usage:

$ curl -i -H "Content-Type: application/json" -X POST -d '{"strings": ["Mario Rossi è nato a Busto Arsizio", "Il signor Di Marzio ha effettuato un pagamento a Matteo", "Marco e Luca sono andati a Magenta"]}' http://localhost:5000/transner/v0.7/ner

$curl -d '{"strings": ["Mario Rossi è nato a Busto Arsizio", "Il signor Di Marzio ha effettuato un pagamento a Matteo", "Marco e Luca sono andati a Magenta"]}' http://localhost:6000/transner/v0.7/ner -H "Content-Type: application/json"

{
  "results": [
    {
      "entities": [
        {
          "offset": 0,
          "type": "PERSON",
          "value": "mario rossi"
        },
        {
          "offset": 21,
          "type": "LOCATION",
          "value": "busto arsizio"
        }
      ],
      "sentence": "Mario Rossi è nato a Busto Arsizio"
    },
    {
      "entities": [
        {
          "offset": 0,
          "type": "PERSON",
          "value": "il signor d'alberto"
        },
        {
          "offset": 49,
          "type": "PERSON",
          "value": "matteo"
        }
      ],
      "sentence": "Il signor D'Alberto ha effettuato un pagamento a Matteo"
    },
    {
      "entities": [
        {
          "offset": 0,
          "type": "PERSON",
          "value": "marco"
        },
        {
          "offset": 8,
          "type": "PERSON",
          "value": "luca"
        },
        {
          "offset": 27,
          "type": "LOCATION",
          "value": "magenta"
        }
      ],
      "sentence": "Marco e Luca sono andati a Magenta"
    }
  ],
  "timestamp": 1581065432.7972977
}

HOW TO USE:

clone the repository and then do:

git submodule init
git submodule update

pretrained models link: https://istitutoboella-my.sharepoint.com/:f:/g/personal/matteo_senese_linksfoundation_com/EvhOF23tja5Nuo3mw03v24oB7D14q9cjk16Ca7xF3nTm-A?e=AWpuiu

conda create --name mediaverse_transner python=3.8 conda activate mediaverse_rest pip install -r requirements.txt

DOCKER

First method

build the image

docker build -t transner-api .

run the image

docker run -d --network host --name mediaverse_transner transner-api

remove container and image

docker ps docker stop mediaverse_transner docker rm mediaverse_transner docker rmi transner-api

Second method

docker-compose up -d

remove container and image

docker-compose down docker rmi transner_web