HeGeL: A Novel Dataset for Geo-Location from Hebrew Text

Paper

The paper can be found here - https://aclanthology.org/2023.findings-acl.460/

Data

The data can be found here - https://github.com/OnlpLab/HeGeL/tree/main/data/human.

The data contains three json files corresponding to three split-sets: train (Tel Aviv), dev (Haifa), and test (Jerusalem).

Each sample contains the following:

• content - place description.
• geometry - the wkt shape of the geolocation of the place.
• goal_point - the centroid of the geometry.

Model

Dependencies

• Pytorch - Machine learning library for Python-related dependencies
• Anaconda - Anaconda includes all the other Python-related dependencies
• ArgParse - Command line parsing in Python

Installation

Below are installation instructions under Anaconda. IMPORTANT: We use python 3.8.15

• Setup a fresh Anaconda environment and install packages:

# create and switch to new anaconda env
$ conda create -n hegel python=3.8.15
$ source activate hegel

# install required packages
$ pip install -r requirements.txt

Instructions

• Here are the instructions to use the code base:

Train and Test Model:

• To train the model with options, use the command line:

$ python train.py --options %(For the details of options)
$ python train.py [-h] [short_name_arg] %(For explanation on the commands)