Skip to content

Latest commit

 

History

History
54 lines (44 loc) · 2.57 KB

README.md

File metadata and controls

54 lines (44 loc) · 2.57 KB

Gnomon Datasets

This repository contains two datasets used to create and evaluate evaluate Gnomon link to be added when camera ready is available.

Please cite our EDBT'25 paper if you use them:

@inproceedings{christodoulakis_2025_gnomon,
  author       = {Christina Christodoulakis and Moshe Gabel and Angela Demke Brown},
  title        = {Metadata Unification in Open Data with Gnomon},
  booktitle    = {Proceedings 28th International Conference on Extending Database Technology,
                  {EDBT} 2025, Barcelona, Spain, March 25-28, 2025},
  year         = {2025}
}

Datasets

Each dataset (dev, test) contains documentation tables extracted from Open Data repositories.

The first line in each file corresponds to the table header, and the following lines correspond to the table data. Headers designate different metadata semantics, which should be mapped to a unified metadata model. Each row corresponds to metadata for a particular data attribute in a different file (not available in this dataset currently).

<dataset>_table_list.csv

This file records the original url the source file of each table was downloaded from.

Ground Truth Mappings

Model

Name Description
dataset The name of the file the described attribute exists in.
name [mandatory] The attribute name, as seen in the data table.
title A human-readable version of the attribute name.
definition A text description of the attribute, may include context.
datatype Specifies which type of value the attribute can have.
scale For numeric attributes, the multiplier (e.g., millions).
format A definition of the structure of data (e.g., ‘YYYY-MM-DD’).
unit The unit of the associated values (e.g., meters, KWh).
key An indication that the attribute is a primary key.
nullable An indication that an attribute value is or isn’t mandatory.
schema Name/URL of an existing schema the attribute belongs to.
examples Sampled values from the attribute domain.
notes Text with miscellaneous extra information

<dataset>_column_mappings.csv

This file records for each table, the mapping between table columns and fields of a flat attribute metadata model.

Attributes:

  • datatable: the id of the csv file containing the table
  • portal: the portal the table was originally sourced from
  • csv_column: the index of a column in that table
  • extracted_name: the name extracted from the header of that column
  • target: the ground truth target mapping to the attribute metadata model.