Repository for SemUN project. It is composed of a docker-compose stack, with:

• An API (un-semun-api)
• A frontend (un-semun-front)
• An NLP pipeline (un-ml-pipeline).
• A Neo4j graph database (neo4j service in docker-compose.yml)
• Scripts to populate the database (un-semun-misc)
• A scraper for the United Nations Digital Library

• 🇺🇳 SemUN repository
  • Table of Contents
    • Description & Paper
    • Running the project
      • Install requirements
      • Run the project
      • Stop the stack
    • Ingest documents using the ML pipeline API

• To have more information on the project, please refer to the project proposal
• For more details about the final result, please refer to the paper

You also need to have Docker installed, I'm using OrbStack as a Docker desktop client for macOS, but regular Docker installation works perfectly fine as well.

When Docker is setup, you just have to run:

# Start the containers
docker-compose up -d

Open the frontend at http://localhost:8080/ if using Docker Desktop or http://un-semun-frontend.un-semun.orb.local/ if using OrbStack.

To stop the stack, just run:

docker-compose down

You are all set! 🎉

To ingest documents, you can use the ML pipeline API. You can find more information about it in the README.md of the un-ml-pipeline folder.

You basically need to send a POST request to the /run endpoint at URL http://un-semun-api.un-semun.orb.local with a JSON body containing the following fields:

[
    {"recordId": "<record_id_0>"},
    {"recordId": "<record_id_1>"},
    {"recordId": "<record_id_2>"},
    ...
]

You can also send a POST request to the /run_search endpoint, at the same URL, with a natural language query to the UN Digital Library. The API will then scrape the results and ingest them in the database.

{
  "q": "<query>"
}

You can also include a limit number of results to scrape, by adding a field "n": <value> in the payload.

For instance:

{
  "q": "Women in peacekeeping",
  "n": 256
}