CC-GPX: Extracting High-Quality Annotated Geospatial Data from Common Crawl

📄 This is the official code repository for the short paper titled 'CC-GPX: Extracting High-Quality Annotated Geospatial Data from Common Crawl', presented at SIGSPATIAL'24 in Atlanta, GA. The pre-print is available on arXiv, and the final paper in the ACM Digital Library.

✍️ Authors: Ilya Ilyankou, Meihui Wang, Stefano Cavazzi, and James Haworth

Abstract

The Common Crawl (CC) corpus is the largest open web crawl dataset containing 9.5+ petabytes of data captured since 2008. The dataset is instrumental in training large language models, and as such it has been studied for (un)desirable content, and distilled for smaller, domain-specific datasets. However, to our knowledge, no research has been dedicated to using CC as a source of annotated geospatial data. In this paper, we introduce an efficient pipeline to extract annotated user-generated tracks from GPX files found in CC, and the resulting multimodal dataset with 1,416 pairings of human-written descriptions and MultiLineString vector data from the 6 most recent CC releases. The dataset can be used to study people's outdoor activity patterns, the way people talk about their outdoor experiences, and for developing trajectory generation or track annotation models.

Setup

We recommend running the notebooks in a separate virtual environment. Using conda,

# Navigate to the project folder
cd cc-gpx

# Create a new virtual environment
conda env create -f environment.yml

# Activate that new virtual environment
conda activate cc-gpx

# Run Jupyter (will open in your default browser)
jupyter lab

Dataset

Run the notebooks in order to build the final GeoPackage dataset with the following fields:

#	Property	Description
1	url	URL of the GPX file
2	warc_file	CC WARC file with GPX file
3	warc_offset	GPX file position in WARC
4	warc_len	GPX file byte length
5	country	Country name as determined by the first point in the track intersecting geoBoundaries
6	desc	Original track description
7	desc_lang	Track description language code, as determined by pycld2
8	desc_en	Track description translated into English
9	elev_source	GPS if elevation is recorded by device; DEM if determined later from Shuttle Radar Topography Mission
10	elev_highest	Track’s highest point, m
11	elev_lowest	Track’s lowest point, m
12	uphill	Cumulative elevation gain, m
13	downhill	Cumulative elevation loss, m
14	length_2d	Track length disregarding elevation, m
15	length_3d	Track length accounting for elevation, m
16	is_circular	True if start and end points are within 350 m from each other, False otherwise
17	geometry	MultiLineString Z geometry in GPS coordinates: (lat, lon, elevation)

Cite

If you find this dataset or workflow useful for your research, please cite us!

@inproceedings{ilyankou2024ccgpx,
      author = {Ilyankou, Ilya and Wang, Meihui and Cavazzi, Stefano and Haworth, James},
      title = {CC-GPX: Extracting High-Quality Annotated Geospatial Data from Common Crawl},
      year = {2024},
      isbn = {9798400711077},
      publisher = {Association for Computing Machinery},
      address = {New York, NY, USA},
      url = {https://doi.org/10.1145/3678717.3691215},
      doi = {10.1145/3678717.3691215},
      booktitle = {Proceedings of the 32nd ACM International Conference on Advances in Geographic Information Systems},
      pages = {693–696},
      numpages = {4},
      keywords = {Common Crawl, GIS, GPS, GPX, hiking, user-generated routes},
      location = {Atlanta, GA, USA},
      series = {SIGSPATIAL '24}
}