This repository is the implemenation of the paper: "A Coherent Unsupervised Model for Toponym Resolution".

• Installation
• Getting Started
• Reference

You need Java 8 (or higher) to use the toponym resolver. All library dependencies are specified in a Maven pom file.

The GeoNames data are stored in SQLite and Redis.

We provided a script, written in Python 3, for Linux machines to lay the groundwork. For a quick start, just run

   python3 install.py

The script downloads GeoNames data, Redis and Apache Maven. After extracting the downloaded files, it prepares a Redis instance by compiling and running it on port 6384 by default.

Once done, the installer runs an importer using Maven to build the SQLite database and initiate the Redis keys. The whole process takes less than 5 minutes to install the requirements and roughly 30 minutes to import the databases (Tested on Ubuntu 14.04 with 4 CPU-cores and 8GB memory).

If you do not need a new Redis instance, specify only the host and port of the instance using the following command:

   python3 install.py --no_redis --redis_port <PORT> --redis_host <HOST> 

This will bypass the Redis installation part.

Here is more details about the arguments for the install script:

    --no_redis        In case no Redis installation is need (not recommended)
    --redis_port      Redis port (default: 6384)
    --redis_host      Redis host (default: localhost)
    --redis_url       Redis URL to download
    --geonames_url    GeoNames data URL to download
    --maven_url       Apache Maven URL to download

The above URLs are provided by default. If the links were broken, you can pass new URLs using the above arguments.

You can create a GeoTagger instance for toponym recognition and resolution. Here is the simplest way to extract toponyms:

GeoTagger geoTagger = new GeoTagger();
List<Toponym> toponyms = geoTagger.extractToponyms("");
for (Toponym toponym : toponyms)
	System.out.printf("%s located at (%.2f, %.2f)\n", toponym.getPhrase(), );

By default, the Context Hierarchy Fusion (CHF) method is employed to resolve toponyms (Refer to the paper for more details).

If you found the code useful, please cite the following paper:

@inproceedings{Kamalloo2018,
 author = {Kamalloo, Ehsan and Rafiei, Davood},
 title = {A Coherent Unsupervised Model for Toponym Resolution},
 booktitle = {Proceedings of the 2018 World Wide Web Conference},
 series = {WWW '18},
 year = {2018},
 isbn = {978-1-4503-5639-8},
 location = {Lyon, France},
 pages = {1287--1296},
 numpages = {10},
 url = {https://doi.org/10.1145/3178876.3186027},
 doi = {10.1145/3178876.3186027},
 acmid = {3186027},
 publisher = {International World Wide Web Conferences Steering Committee},
 address = {Republic and Canton of Geneva, Switzerland},
 keywords = {context-bound hypotheses, geolocation extraction, spatial hierarchies, toponym resolution, unsupervised disambiguation},
}