=Paper=
{{Paper
|id=Vol-2038/paper7
|storemode=property
|title=Repolytics: Identifying Measurable Insights for Digital Repositories
|pdfUrl=https://ceur-ws.org/Vol-2038/paper7.pdf
|volume=Vol-2038
|authors=Vangelis Nomikos
|dblpUrl=https://dblp.org/rec/conf/ercimdl/Nomikos17
}}
==Repolytics: Identifying Measurable Insights for Digital Repositories==
https://ceur-ws.org/Vol-2038/paper7.pdf
Repolytics: Identifying Measurable Insights for Digital
Repositories
Vangelis Nomikos
Intelibility Innovative Data Engineering LLC,
Atlanta, GA, USA
nomikos@intelibility.com
Abstract. This paper presents specific aspects of the Repolytics platform: a da-
ta analytics platform for digital repositories. An overview of the platform is pre-
sented alongside an example of how one can employ it's services to analyze a
digital repository's data and identify quality related issues.
Keywords: metadata quality, data quality, data analytics, quality metrics, digi-
tal repositories.
1 Introduction
Data is captured, analyzed and used to drive all aspects of our lives in a data driven
world. Recently, we have seen a rapid growth of digital repositories and open data
catalogues being made available to the public. In the case of digital repositories that
target the libraries-archives, scientific-research data domains and open data portals the
market is dominated by open source solutions such as DSpace, Omeka, DKAN,
CKAN due to their simplicity and low-no cost. Most solutions enable ingest / cata-
loguing of information either through automated means (SWORD, REST APIs) or
through simple and untuitive forms. Quality assurance in most cases comprises of a
number of mandatory metadata the user has to enter. The reality however is far more
complex and this has a profound effect in the quality of the the ingested data.
Repolytics [1] is a platform that aims at filling this gap through intelligent data
analytics. The data loaded into the platform are analyzed and specific quality metrics
are presented alongside a more thorough analysis per metadata element. The metrics
include metadata completeness, accuracy and consistency. Furthermore, similar met-
rics are calculated for the data as well.
2 Related Work
Metadata quality is an important issue for the Digital Library domain and has attract-
ed several researcher groups to deal with it. One of the main demands is the estab-
lishment of a conceptual framework consisted of a set of well-defined quality assess-
ment criteria such as completeness, validity, consistency, timeliness, appropriateness
and accuracy constituents [2, 3]. In such a framework any assessment effort would be
based on reliable indications about metadata quality. The first attempt to define a
framework established a narrow set of criteria such as accuracy, completeness and
2
serviceability [4]. Some researches expanded the criteria set from which metadata
quality is approached [6, 8], while some other narrowed the perspective focusing only
on the completeness criterion [5].
An important evolution of this scientific field was the introduction of the context-
dependent metadata quality approach [3, 7]. According to this approach metadata
quality issues follow four major concepts: mappings, changes to the information enti-
ty, changes to the underlying entity and context changes. For these concepts a taxon-
omy of 22 information quality criteria was developed. The criteria were clustered to
three categories: intrinsic, relation and reputational and are measured via 41 metrics
[7]. Recent research suggests that a metadata quality framework doesn’t have to “in-
vent new dimensions in order to accommodate the needs of diverse communities of
practice” [3] but to give the flexibility to each evaluator to assess the results within a
specific context. This paper follows the context-dependent approach and assumes that
metadata quality strongly depends on the viewpoint of the evaluator and should be
aligned with the application domain for which the metadata were produced and used.
Therefore weighting functions for these factors (evaluator viewpoint, application do-
main, metadata usage) should be defined and used to weight the values of the metada-
ta quality metrics.
3 The Repolytics approach
The core principle of the Repolytics platform is the provisioning of an expandable set
of middleware services that operates both on metadata and data level of a digital re-
pository. The most fundamental services of the platform involve:
(meta)-data integration services
(meta)-data profiling services
(meta)-data quality services
The Repolytics platform enables the user to load data from different data sources ei-
ther directly (e.g. through a file archive) or by using one of the supported data provid-
ers such as DKAN API and OAI-PMH protocol. Each data source provides one or
more metadata (e.g. OAI-DC, MODS) and data (e.g. CSV, excel) representations.
Once the data has been loaded, each digital object is analyzed and the main workflow
seen in Figure 1 below is executed.
Figure 1. Overal workflow per digital object.
For each digital object, if metadata is provided, the format is identified, every metada-
ta element is indexed and a set of core analyzers are invoked for each element. For
specific elements depending on their type, a set of custom analyzers is invoked. If the
3
data is provided (usually throuh some kind of URL) the actual data is fetched and
verified followed by a profile of each bitstream. Again, a set of custom analyzers is
then invoked depending on the file type.
Some of the metrics that are calculated include:
metadata completeness
distinct values for each metadata element
accuracy for specific element types (e.g. dates, actors) where all values are
classified according to their class
itemset frequencies when applicable (e.g. in the case of subject terms and
keywords)
One of the primary challenges include the efficeint visualization of the results to the
end user. For that reason a series of bar charts, radar charts, tables, gauge meters etc
are employed per case. For example, as shown in Figure 2 a radar chart is employed
to fingerprint an entire repository according to it's completeness.
Figure 2. Radar chart showing the overall completeness for a data soure.
Similary, an accuracy detector identifies and classifies all metadata element values
according to their class and provides an insight on the accuracy level of each class
(low, medium, high).
Class Type Example Accuracy
YYYY-MM-DD 2015-12-20 Medium
YYYY 1998 Low
YYYY-MM-DD 1996-02-23 14:30:11 High
HH:MM:SS
UNKNOWN Sp.1996/23 -
4
4 Conclusions & Future Work
In this paper, Repolytics, a platform for repository data analytics is presented. The
platform enables the use to easily load data from supported data sources and analyzes
this data focusing primarily on completeness and accuracy whereas in specific cases,
more specific metrics (such as itemset frequencies) are employed to help gain insight
on highly subjective metrics such as consistency. The platform can also access and
profile data as well.
References
1. Repolytics Platform, http://www.repolytics.com
2. Herzog, T., Scheuren, F., Winkler, W.: Data Quality and Record Linkage Techniques.
Springer-Verlag, NY (2007)
3. Tani, A., Candela, L., Castelli, D.: Dealing with metadata quality: The legacy of digital li-
brary efforts. Inf. Process. Manag. 49 (2013) 1194–1205
4. Moen, W.E., Stewart, E.L., McClure, C.R.: The Role of Content Analysis in Evaluating
Metadata for the U.S. Government Information Locator Service (GILS): Results from an
Exploratory Study. http://digital.library.unt.edu/ark:/67531/metadc36312/citation/ (1997)
5. Margaritopoulos, M., Margaritopoulos, T., Mavridis, I., Manitsaris, A.: Quantifying and
measuring metadata completeness. J. Am. Soc. Inf. Sci. Tech. 63 (2012) 724–737
6. Ochoa,X., Duval,E.: Quality Metrics for Learning Object Metadata. In: WorldConference
on Educational Multimedia, Hypermedia and Telecommunications. (2006) 1004–1011
7. Stvilia, B., Gasser, L., Twidale, M.B., Smith, L.C.:A framework for information quality
assessment. J. Am. Soc. Inf. Sci. Tech. 58 (2007) 1720–1733
8. Chen,Y.-N.,Wen,C.-Y.,Chen,H.-P.,Lin,Y.-H.,Sum,H.-C.:Metrics for metadata quality as-
surance and their implications for digital libraries. In: Xing, C., Crestani, F., and Rauber,
A. (eds.) 13th International Conference on Asia-Pacific Digital Libraries, Beijing, China.
Springer-Verlag, Berlin (2011)138–147