Insights based on data are omnipresent. However, in particular in modern data analytics applications, information about the underlying data often remain obscure, hindering accountable data analytics.
Recent efforts have been put into better describing such data based on
metadata, similarly to what has been done in various scientific disciplines for transparent and reproducible research. Based on a detailed
study of various metadata standards and proposals, we observe that existing metadata models do not yet sufficiently cover information that
is relevant for data accountability. To fill this gap, this paper proposes
LiQuID, a novel metadata model to make datasets accountable throughout their life cycle. It is more general than existing metadata models,
which can be mapped to LiQuID. We validate LiQuID for the purpose
of dataset accountability based on a real-world workload we created.
%0 Journal Article
%1 oppold2020accountable
%A Oppold, Sarah
%A Herschel, Melanie
%D 2020
%K ethischeFragestellungen metadata quality rechtlicheFragestellungen
%T Accountable Data Analytics Start with Accountable Data: The LiQuID Metadata Model
%X Insights based on data are omnipresent. However, in particular in modern data analytics applications, information about the underlying data often remain obscure, hindering accountable data analytics.
Recent efforts have been put into better describing such data based on
metadata, similarly to what has been done in various scientific disciplines for transparent and reproducible research. Based on a detailed
study of various metadata standards and proposals, we observe that existing metadata models do not yet sufficiently cover information that
is relevant for data accountability. To fill this gap, this paper proposes
LiQuID, a novel metadata model to make datasets accountable throughout their life cycle. It is more general than existing metadata models,
which can be mapped to LiQuID. We validate LiQuID for the purpose
of dataset accountability based on a real-world workload we created.
@article{oppold2020accountable,
abstract = {Insights based on data are omnipresent. However, in particular in modern data analytics applications, information about the underlying data often remain obscure, hindering accountable data analytics.
Recent efforts have been put into better describing such data based on
metadata, similarly to what has been done in various scientific disciplines for transparent and reproducible research. Based on a detailed
study of various metadata standards and proposals, we observe that existing metadata models do not yet sufficiently cover information that
is relevant for data accountability. To fill this gap, this paper proposes
LiQuID, a novel metadata model to make datasets accountable throughout their life cycle. It is more general than existing metadata models,
which can be mapped to LiQuID. We validate LiQuID for the purpose
of dataset accountability based on a real-world workload we created.},
added-at = {2021-01-15T15:53:15.000+0100},
author = {Oppold, Sarah and Herschel, Melanie},
biburl = {https://puma.ub.uni-stuttgart.de/bibtex/2b399927cecc700a8d82e28f0b6304780/diglezakis},
interhash = {ed7235584594c29440adfe8a80e92cbd},
intrahash = {b399927cecc700a8d82e28f0b6304780},
keywords = {ethischeFragestellungen metadata quality rechtlicheFragestellungen},
timestamp = {2021-01-15T14:53:15.000+0100},
title = {Accountable Data Analytics Start with Accountable Data: The LiQuID Metadata Model},
year = 2020
}
