This repository provides a Dockerized version of Mordecai 3, a powerful neural geoparser and event geocoder developed by Andrew Halterman. Mordecai 3 leverages state-of-the-art techniques to accurately extract geographical information from text, making it an invaluable tool for location-based data analysis.
The core geoparsing engine, Mordecai 3, is the creation of Andrew Halterman. For more details on the model and its accuracy, refer to the Mordecai 3 paper.
If you use Mordecai 3 in your work, please give proper credit to the developer by citing the following paper:
@article{halterman2023mordecai,
title={Mordecai 3: A Neural Geoparser and Event Geocoder},
author={Andrew Halterman},
year={2023},
journal={arXiv preprint arXiv:2303.13675}
}- Build the geonames index as described in https://github.com/openeventdata/es-geonames Alternatively, you can download a pre-built geonames index here. Extract the folder and save it in project directory.
- Start the service with the following command
docker compose up
- After that the API endpoint is accessible under http://localhost:8080/geoparse/
The following code shows an example how make request to the mordecai 3 API in python
import requests
item = {
"page": "test",
"title": "test title",
"time": "test time",
"author": "test author",
"content": "Austria & Germany are both really beautiful countries"
}
URL = 'http://localhost:8080/geoparse/'
response = requests.post(URL, json=item)
results = response.json()The output should look like this
{
"page": "test",
"title": "test title",
"time": "test time",
"author": "test author",
"content": "Austria & Germany are both really beautiful countries",
"location_information": [
{
"geonameid": "2782113",
"placename": "Republic of Austria",
"longitude": 13.33333,
"latitude": 47.33333
},
{
"geonameid": "2921044",
"placename": "Federal Republic of Germany",
"longitude": 10.5,
"latitude": 51.5
}
]
}