text-contextifyer

Turn plain Markdown into enriched Markdown with ontology-based hyperlinks. This tool supports small-medium sized ontologies only right now, as they need to be loaded into memory.

Features

• Load an RDF/OWL ontology into memory from a SPARQL endpoint.
• Extract labels (rdfs:label, skos:prefLabel)
• Match words in Markdown text against ontology terms (fuzzy or exact)
• Replace matches with hyperlinks in the markdown file supplied

Example

Input:

Computer science and Geology are fascinating fields.

Output:

[Computer science](someurl-about-computerscience.org) and [geology](some-otherurl-related-to-geology.org) are fascinating fields.

Usage

First, make sure your GraphDB SPARQL endpoint is running and create a .env file based on .env.dist with your configuration.

Running Locally

To run tests:

poetry install
pytest

To start the microservice locally:

PYTHONPATH=src poetry run uvicorn text_contextifyer.api.main:app --reload

Running with Docker

1. Build the Docker image:

docker build -t text-contextifyer .

2. Run the container (choose one of the following methods):

   a. If GraphDB is running on your host machine:

   docker run --rm --network=host --env-file .env text-contextifyer

   b. Or using port mapping and Docker's host resolution:

   # First, modify your .env file to use host.docker.internal instead of localhost:
   # ONTOLOGY_SPARQL_ENDPOINT=http://host.docker.internal:7200/repositories/your-repo
   docker run --rm -p 8000:8000 --env-file .env text-contextifyer

The API documentation will be available at http://localhost:8000/docs

Testing the API

Once the service is running, you can test it with curl (make the ontology you point to contains labels that appear in the text you are contextifying):

curl -X POST http://localhost:8000/contextify \
  -H "Content-Type: application/json" \
  -d '{"markdown":"Computer science and Geology are fascinating fields."}'