gdelttools 0.7b2

Loading GDELT data into MongoDB

This is a set of programs for loading the GDELT 2.0 data set into MongoDB.

Quick Start

Install the latest version of Python from python.org. You need at least version 3.6 for this program. Many versions of Python that come pre-installed are only 2.7. This version will not work.

Now install gdelttools

pip install gdelttools

Now get the master file of all the GDELT files.

gdeltloader --master

This will generate a file named something like masterfilelist.txt

Downloading the master data set

To download the master data set associated with GDELT (the export files) you can combine these steps:

gdeltloader --master --download --overwrite

This will get the master file, parse it, extract the list of CSV files and unzip them. the full GDELT 2.0 database runs to several terabytes of data so this is not recommend.

The overwrite argument ruthlessly overwrites all files with extreme prejudice. Without it the gdeltloader script will attempt to reuse the files you have already downloaded. As each file is unique this may save time if you need to re-download some files.

To limit the amount you download you can specify --last to define how many files worth of data you want to download:

gdeltloader --master --download --overwrite --last 20

This command will download the most recent 20 files worth of data. Note that a file is a triplet of export, mentions and gkg data. If you only want one you should specify a --filter. Without the filter a command like the above will actually download 60 files.

GDELT 2.0 Encoding and Structure

The GDELT dataset is a large dataset of news events that is updated in real-time. GDELT stands for Global Database of Events Location and Tone. The format of records in a GDELT data is defined by the GDELT 2.0 Codebook

Each record uses an encoding method called CAMEO coding which is defined by the CAMEO Codebook.

Once you understand the GDELT recording structure and the CAMEO encoding you will be able to decode a record. To fully decode a record you may need the TABARI dictionaries from which the CAMEO encoding is derived.

How to download GDELT 2.0 data

The gdeltloader script can download cameo data and unzip the files so that they can be loaded into MongoDB.

usage: gdeltloader [-h] [--master] [--update] [--database DATABASE] [--collection COLLECTION]
  [--local LOCAL] [--overwrite] [--download] [--metadata]
  [--filter {all,gkg,mentions,export}] [--last LAST] [--version]

options:
  -h, --help            show this help message and exit
  --master              GDELT master file [False]
  --update              GDELT update file [False]
  --database DATABASE   Default database for loading [GDELT]
  --collection COLLECTION
                        Default collection for loading [events_csv]
  --local LOCAL         load data from local list of zips
  --overwrite           Overwrite files when they exist already
  --download            download zip files from master or local file
  --metadata            grab meta data files
  --filter {all,gkg,mentions,export}
                        download a subset of the data, the default is all data [export, mentions gkg, all]
  --last LAST           how many recent files to download default : [0] implies all files
  --version             show program's version number and exit

Version: 0.07b1 More info : https://github.com/jdrumgoole/gdelttools

Here is how to download the last 5 hours of GDELT data.

gdeltloader --master --update --download --last 20

This command will only download the export files for the last 20 15-minute blocks, which are the files we are interested in.

How to import downloaded data into MongoDB

Now import the CSV files with mongoimport.

There is a mongoimport.sh script in the gdelttools repo which is already configured with the right arguments. There is also a corresponding field file, gdelt_field_file.ff which this script uses to ensure correct type mappings.

To run:

sh mongoimport.sh --uri "<YOUR-MONGODB-CONNECTION-STRING>"

This will upload all the CSV files in the current working directory.