Spatial discovery and the research library

Abstract

Academic libraries have always supported research across disciplines by integrating access to diverse contents and resources. They now have the opportunity to reinvent their role in facilitating interdisciplinary work by offering researchers new ways of sharing, curating, discovering, and linking research data. Spatial data and metadata support this process because location often integrates disciplinary perspectives, enabling researchers to make their own research data more discoverable, to discover data of other researchers, and to integrate data from multiple sources. The Center for Spatial Studies at the University of California, Santa Barbara (UCSB) and the UCSB Library are undertaking joint research to better enable the discovery of research data and publications. The research addresses the question of how to spatially enable data discovery in a setting that allows for mapping and analysis in a GIS while connecting the data to publications about them. It suggests a framework for an integrated data discovery mechanism and shows how publications may be linked to associated data sets exposed either directly or through metadata on Esri's Open Data platform. The results demonstrate a simple form of linking data to publications through spatially referenced metadata and persistent identifiers. This linking adds value to research products and increases their discoverability across disciplinary boundaries.

Full text

UC Santa Barbara
UC Santa Barbara Previously Published Works
Title
Spatial discovery and the research library
Permalink
https://escholarship.org/uc/item/8g3714v0
Journal
Transactions in GIS, 20(3)
Authors
Lafia, Sara K
Kuhn, Werner
Publication Date
2016-06-28
Data Availability
The data associated with this publication are available at:
http://discovery.ucsb.opendata.arcgis.com
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California

1
Spatial Discovery and the Research Library
Sara Lafia1, Jon Jablonski2, Werner Kuhn1, Savannah Cooley1, F. Antonio Medrano1
1 Center for Spatial Studies
3512 Phelps Hall
Department of Geography
1832 Ellison Hall
Department of Geography
University of California at Santa Barbara
Santa Barbara, CA 93106-4060
2 UC Santa Barbara Library
University of California at Santa Barbara
Keywords
digital libraries, spatial data portals, Semantic Web

2
Abstract
Academic libraries have always supported research across disciplines by integrating access to diverse
contents and resources. They now have the opportunity to reinvent their role in facilitating
interdisciplinary work by offering researchers new ways of sharing, curating, discovering, and linking
research data. Spatial data and metadata support this process because location often integrates disciplinary
perspectives, enabling researchers to make their own research data more discoverable, to discover data of
other researchers, and to integrate data from multiple sources. The Center for Spatial Studies at the
University of California, Santa Barbara (UCSB) and the UCSB Library are undertaking joint research to
better enable the discovery of research data and publications. The research addresses the question of how
to spatially enable data discovery in a setting that allows for mapping and analysis in a GIS while
connecting the data to publications about them. It suggests a framework for an integrated data discovery
mechanism and shows how publications may be linked to associated data sets exposed either directly or
through metadata on Esri’s Open Data platform. The results demonstrate a simple form of linking data to
publications through spatially referenced metadata and persistent identifiers. This linking adds value to
research products and increases their discoverability across disciplinary boundaries.

3
1. Introduction
Location plays a key role in the organization and integration of knowledge. In an interdisciplinary setting,
location can reveal patterns and trends in diverse and seemingly disparate information. For example, a
“geographic prism” on social mobility data in the United States reveals vast regional differences that can
then produce hypotheses about causes, based on local differences in factors like family structure or
schools (“Mobility, measured”, The Economist 2014). Data discovery tools that exploit location can offer
users a spatial view of phenomena, and in doing so, bridge disciplines in research and policy making. The
design of such tools, with an emphasis on connecting the discovered data to publications about them, is
the focus of this paper.
1.1. Problem statement
Enabling the spatial discovery of research publications and datasets, herein referenced as research objects,
is the next step in the evolving role of the modern research library. The notion of the extensible and
reusable research object originates from the domain of e-Science (Bechhofer et al. 2010). Over time, the
set of research objects, beginning with documents, has expanded beyond texts to include artifacts, models,
games, and works of art (Buckland 1997). Today, research libraries are increasingly called upon to build
links between research objects, such as journal articles or electronic theses, and auxiliary data, of which
both may have embedded locational references and may reside in external data repositories.
Emerging research object repositories, which hold data and publications, are still unstable as
architectures and face challenges in handling spatially referenced content (Hey et al. 2009). There is a
growing need for a stable yet flexible discovery mechanism that can thrive in an evolving spatial and non-
spatial information landscape (Cooley et al. 2015). At the same time, e-Science is producing sophisticated
models of research objects (Bechhofer et al. 2010) that are exceedingly complex for the needs of data and
publication discovery at libraries. Much work remains to be done in the development of simple search and
discovery tools that span multiple collections (van Hoolen et al. 2014).

4
In this article, we address the primary challenge of stability by implementing a simple linked data
model that exploits basic relationships between research data and research publications in a way that does
not break when repositories change. In doing so, we also address a second challenge, that of supporting
discovery, resulting in enhanced integrative capacity for spatially referenced research objects. Combining
these two challenges in the proposed pragmatic form is expected to result in progress on a third, broader
goal: that of supporting interdisciplinarity in scientific workflows through data reusability, within and
across domains. These three challenges translate into the following set of guiding research questions:
1. How can libraries generate stable links for research objects across repositories?
2. How can libraries support the discovery of research objects based on location?
3. How can libraries promote cross-disciplinary data sharing and reuse?
The research, undertaken by the Center for Spatial Studies at the University of California, Santa Barbara
(UCSB), in partnership with the UCSB Library and Esri Inc., seeks to make spatial references and
relationships explicit in research objects, thereby integrating diverse contents and contextualizing
published research data by connecting them to publications.
1.2. Motivation
Research institutions generate massive quantities of data from diverse disciplines in a wide variety of
formats (Mayernik et al. 2015). Recent efforts to increase transparency and reproducibility encourage, and
often mandate, that researchers make publications and data publically accessible through open-access
licenses (University of California Regents 2014). A proliferation of associated data is contributing to a
growing imbalance between an institution’s ability to collect data and its ability to curate resources
(Cragin et al. 2010), resulting in a trade-off between quality assurance and ingestion capacity, a trend that
the UCSB Library can attest has accelerated in the intervening years.
Interdisciplinary research presents additional unique challenges, including disciplinary
differences in frames of reference, operational agendas, research methods, and vocabularies (Brewer
2015; MacMillan 2014). Many data discovery portals support only domain-specific vocabularies, data

5
structures, and metadata formats, severely limiting the applicability and reuse of data across domains
(Golding 2009). More limitations arise when subsets of domain research are published in expensive
subscription journals. This diminishes research impact and potential for data reuse across domains, which
could be enhanced if made available through open-access policies (Harnad et al. 2008).
Library repositories have developed detailed workflows for generating metadata that use rigorous
metadata content standards and controlled vocabularies, but result in extremely limited ingest capacity. In
contrast, repositories for self-deposit of spatial content, such as ArcGIS Online,1 have minimal metadata
constraints and see 8,000–12,000 new and mostly undescribed objects added per day (Szukalski 2015).
Metadata that describe the lifecycle of a dataset are often very granular and must account for both library
and spatial needs. Metadata are valuable for long-term preservation, yet they are not central to resource
discoverability (Hardy and Durante 2014), which is the primary focus of this research. Enforcing
particular metadata requirements may do more to hinder data availability, especially across diverse
domains that have their own metadata standards, than to aid in their discovery.
Library-run and self-deposit systems of research data management can be complementary, as they
approach control and sharing in two distinct ways. However, they are not currently connected. This
research proposes to align the traditional library ingest process with the self-deposit approach of cloud-
based GIS, such as ArcGIS Online, through the generation of links between two sets of research objects:
researcher publications and researcher data. This approach combines the best aspects of both worlds:
spatial discovery of data from the GIS world and document curation from the library world, connected
through a lightweight and stable linked data solution.
Creating links between publications held in a tightly controlled library repository and data stored
across external databases increases the discoverability of these research objects. This work connects
research objects held in separate repositories without the need to formally align their metadata schemas.
In our proposed framework, a research object in a self-deposit environment like ArcGIS Online, which
1 https://www.arcgis.com/home/

6
has minimal metadata constraints, is semantically linked to a related research object in an institutional
environment with tightly controlled metadata.
This article presents a proof-of-concept model for linking spatial data to the research publications
that utilize them. Using OpenRefine with its Resource Description Framework (RDF) extension for data
processing and cleaning2, we link sample publications to data hosted on Esri’s Open Data platform by
Dublin Core metadata relationships. The linked data are stored as triples, which allows for queries on the
associated RDF about publication data. Such formalized relationships are key to developing a rich
publication and data repository that allows for discovery of research resources and advances cross-
disciplinary sharing of knowledge, as illustrated in Figure 1.
Figure 1—Project vision for data discovery and publication integration across domains
2. Background and related work
This work builds on a long tradition of spatially enabled digital libraries and uses the latest semantic and
geospatial technologies to demonstrate the potential for spatial discovery and the interlinking of research
resources. As university researchers are increasingly expected to share the data associated with their
publications under open data mandates, university libraries find themselves being called upon to curate
increasing volumes and additional types of researcher-generated data. In this context, enhancing users’
ability to share, discover, and make sense of content is of great importance.
2.1. Library repositories
In the mid-1990s, the Alexandria Digital Library (ADL) at UCSB was the first distributed digital library
(Freeston 2004) to offer collections of georeferenced materials, hosted online, searchable by spatial and
temporal criteria (Goodchild 2004). ADL eventually lapsed, relegating UCSB researcher data, such as the
popular Maya Forest GIS collection (Ford 1995), to offline discovery and curation. Reinstating such
legacy collections through an open-access digital presence increases their utility in an interdisciplinary
2 http://openrefine.org/download.html

7
research context. Further, linking these datasets to publications in a manner that can be exploited by
Semantic Web tools improves their discoverability.
Many university libraries have implemented hybrid ad-hoc solutions for spatial data collection,
discovery, access, storage, and archiving in the context of the changing landscape of user needs and
technologies (Scaramozzino et al. 2014). Libraries have generally promoted interdisciplinary
collaboration by supporting geospatial research platforms and tools for analysis and post-data discovery.
However, they do not yet combine spatial and semantic approaches to expose connections between
existing data silos that span diverse disciplines. In practice, most library-curated research objects are
locally stored, have limited access points, and are undiscoverable from related content (Padilla 2016).
UCSB Map & Imagery Laboratory (MIL) is in the process of developing a spatial metadata
workflow using ArcCatalog for the purposes of preparing spatial datasets for ingest into the new
Alexandria Digital Research Library (ADRL)3. The ISO 19115 standard4, the Open Geoportal Metadata
Creation Guide5 and the Stanford University metadata creation workflow6 inform this metadata model.
The work described in this paper couples these ongoing efforts with the production of linked data.
2.2. Emerging spatial data technologies
Achieving the dual purposes of enhancing spatial discovery and linking research objects requires a novel
solution. Some contemporary data management solutions address the need to enable the spatial discovery
of resources, but do not enhance discovery of resources through semantic links. GeoBlacklight7, for
instance, is an open source, multi-institutional software project that many libraries are currently adopting
(Addison et al. 2015; Durante and Hardy 2015). It offers users text-based, spatial, and faceted semantic
search to enable discovery of GIS-consumable resources across organizations (Hardy and Durante 2014).
GeoBlacklight also allows users to connect to data as a service, which enables analysis from a desktop
3 http://alexandria.ucsb.edu/
4 http://www.iso.org/iso/catalogue_detail.htm?csnumber=53798
5 http://opengeoportal.org/working-groups/metadata/metadata-creation-guide/
6 https://lib.stanford.edu/metadata/documentation
7 https://github.com/geoblacklight/geoblacklight

8
GIS, comparable with Esri Open Data. While a GeoBlacklight instance for the UCSB Library would
support spatial discovery by relating spatially referenced content based on location, it would not connect
the data to publications held in the library’s own repositories, nor to other external repositories. Another
possible data management solution includes the California Digital Library’s Dash system, which has been
adopted by several University of California campuses and features a self-deposit feature, facilitates data
search, data sharing, and preservation services (Tsang 2015). However, DASH does not offer inherent
spatial functionality, although efforts to achieve this are underway at UC Irvine8.
Considering these existing alternatives, utilizing Esri’s ArcGIS Online platform as a foundation
for combined spatial and semantic search makes sense for several reasons. Since GIS software has
become ubiquitous for performing spatial analysis across a variety of academic disciplines, universities
often administer an ArcGIS Online enterprise account through their libraries. ArcGIS Online is a cloud-
based GIS that acts as a self-deposit data system with basic geoprocessing functionality. Additionally,
ArcGIS Online now includes Esri Open Data9, which is a spatial data repository with native access
controls and search features. Enabling Open Data on ArcGIS Online allows organizations to make content
available to the public or restricted to users authorized by the institution.
There are many advantages to using Esri Open Data as a spatial data discovery solution, not the
least of which is publishing spatial data in a way that allows open access and download. Users are not
required to have ArcGIS Online credentials to access data hosted through Esri Open Data, which
increases both accessibility to data and reproducibility of results derived from that data. ArcGIS Online
also supports various metadata standards, increasing the potential to share data across domains. Its
interface allows for visualization and filtering of the data for basic geoprocessing and analysis. This adds
immediate value to the discovery process, as users can begin making sense of datasets even before
8 https://dash.lib.uci.edu/xtf/search
9 http://opendata.arcgis.com/

9
downloading them. Many organizations are adopting Esri’s Open Data platform because ArcGIS Online
offers web-based analysis and search. UCSB’s instance of Esri Open Data10 is shown in Figure 2.
Figure 2—UCSB’s Open Data instance leverages ArcGIS Online
While Esri’s Open Data platform is an excellent tool for publishing, discovering, and accessing
spatial datasets, it is not a stable repository solution in either the traditional sense of institutional preprint
repositories or in the emerging sense of Trusted Digital Repositories (Tsang 2015). However, when
linking data with a controlled resource, such as UCSB’s Alexandria Digital Research Library11 (ADRL)
repository, which hosts theses and dissertations, or University of California’s eScholarship12, which offers
open-access to researcher publications, the power of the Semantic Web can be brought to bear on the
systems. This design choice provides flexibility that many current repositories cannot offer.
2.3. State of the art
Many institutions, including libraries, archives and museums, are adopting linked data approaches to
improve the discoverability of the growing number of resources that they curate (van Hooland et al.
2014). Institutions, such as the Linked Data for Libraries university consortium, the Library of Congress
and the Tate Modern Gallery, leverage linked open data technologies to enhance access to their
collections. These management models offer users access to data and metadata through Application
Programming Interfaces (APIs) and extend query capabilities through accessible endpoints.
The Linked Data for Libraries (LD4L) 13 initiative is a multi-institutional effort, including
Stanford, Rice and Harvard universities, aimed toward applying the Library of Congress Bibliographic
Framework Initiative to describe library resources. Transforming traditional MARC (MAchine-Readable
Cataloging) metadata descriptions, which are flat, text-based, and fielded (Avram 2003) into linked
BIBFRAME descriptions for cartographic and geospatial materials leverages Library of Congress
controlled vocabularies alongside DBPedia and GeoNames, to model places, creators, themes, and events
10 http://discovery.ucsb.opendata.arcgis.com/
11 http://www.alexandria.ucsb.edu/
12 http://escholarship.org/
13 https://www.ld4l.org/

10
(Durante et al. 2016). Library of Congress is a notable early organizational contributor to the production
of API-accessible linked open data for authority files14. Many institutions use these services for
authoritative reconciliation (Heath and Bizer 2011). These services allow institutions, such as the Tate
Modern Gallery, to contribute collection metadata to repositories, like GitHub15, that they neither own nor
manage, increasing discoverability, content exposure and creative reuse (Padilla 2016).
The development of Semantic Web technologies enables linked data driven portals. Linked data
portals provide new opportunities to organize metadata and retrieve information resources such as text
documents, datasets, and multimedia content (Baierer et al. 2014; Hu et al. 2015-1; Hu et al. 2015-2).
Linked data resource discovery systems can index domain-specific information with terms from
ontologies. Ontologies are formal explicit specifications of a shared conceptualization using a vocabulary
of classes and relations, expressed in RDF, which is a data model that stores metadata attributes as nodes
and links to constitute an interconnected graph.
Whereas other methods for publishing data rely on multiple data models, the RDF data model
provides an integrated and simple access mechanism that also supports hyperlink-based data discovery
using uniform resource identifiers (URIs) as global identifiers for entities (Heath and Bizer 2011). For
instance, Athanasis et al. (2009) described data with domain-specific spatial ontologies in a linked data
discovery tool, and Keßler et al. (2012) developed a linked data portal for the GIScience community to
explore and visualize geographic distributions of publications by conference location and editor or author
affiliations. Scheider et al. (2014) have leveraged linked spatiotemporal data to enhance access to diverse
formats of library materials, from paper maps to scientific datasets. Taken together, the interlinking of
research objects and their metadata creates a semantically linked graph.
Adopting semantic technologies addresses issues of interoperability that arise from online portals
featuring spatial data in various standards and formats. In particular, relationships between research
14 http://id.loc.gov/
15 https://github.com/tategallery/collection

11
publications and associated data can be captured through RDF subject-predicate-object triples, which
bridge gaps between data and metadata, as well as differing metadata content standards.
3. Methods
Publicly available research objects, namely researcher datasets and researcher publications, drive the data
discovery mechanism developed in this research. The design and evaluation of a linked data model is
informed by user personas, which structure the relationship between published research and associated
data. The extensible triple model developed in this work allows for future expansion of the vocabulary.
3.1. User personas
The current designs of most access systems do not support the spatial integration of research object
collections across various domains. Adopting the personas of domain scientists and considering the types
of data that each might search for or contribute, along with their motivations for doing so, informed the
design specifications of our system. The UCSB Esri Open Data instance contains collections of test data
that span research domains, data formats, and user needs. The current three exemplary data collections
represent a small but diverse range of disciplines, from archaeology to political science, and diverse
formats, including shapefiles, imagery, text documents, external repositories, and map services.
Table 1—Personas, domains, and datasets of researchers currently discoverable through UCSB Open Data
The first collection corresponds to Anabel Ford’s Maya Forest GIS and was obtained from a CD archive
(Ford 1995). The data include shapefiles and imagery complete with full ISO compliant metadata created
by UCSB Library staff. The second collection comes from a meta-analysis conducted by Benjamin
Halpern, a UCSB ecologist. His collection of sampling sites has a global extent and is hosted in an
external repository, a practice typical of UCSB researchers in the life sciences for disseminating research
(Halpern et al. 2009). While these spatial data are open-access and publically shared, they are not
currently discoverable through a search of UCSB Library holdings. The third data collection comes from
Thomas Patterson, a political scientist at Stanford University, and represents world boundaries of disputed

12
areas. The data are part of the broader Natural Earth collection, currently discoverable through the UCSB
Library, but not yet formally associated with Patterson’s research publications (Patterson 2009).
The personas cover various data sharing scenarios. Anabel Ford has locally hosted resources that
she intends to share with a global public audience through open access. Benjamin Halpern is from a
domain that favors data distribution through a repository external to UCSB. Thomas Patterson is from
another institution and has spatially relevant contents that might interest Ford, Halpern, or other scientists
at UCSB or anywhere else. The datasets share a spatial overlap that would not otherwise be obvious. For
instance, Patterson’s contested borders dataset is a feature collection with a global extent, yet intersects
with Ford and Halperns’ regions of research. The potential to expose the spatial complementarity of
resources would go unrecognized without the assignment of spatial footprints to these objects. Users can
then benefit from discovering useful and seemingly unrelated datasets or publications from unfamiliar
domains by exploring the spatial relations of the research objects.
Taken together, these exemplary researcher personas and associated datasets provide a foundation
for several competency questions that capture the kinds of queries that users may want to construct:
● Find datasets referenced by a particular publication.
● Find publications that have a particular dataset associated with them.
● Find research objects that overlap with a particular spatial extent.
Performing such queries is frequently relevant to a resource discovery process, but relationships between
research data, the publications that reference them, and the locational extents that they cover are not
currently exposed in the metadata. The onus of relating publications with datasets, as well as relating both
the publications and datasets with location, is currently placed on the end-user. The linked data
relationship between research publications and data, taken along with the spatial extent of the dataset
represented in Open Data, address the types of thematic and spatial queries that users would currently like
to ask of a library catalog but cannot.
3.2. Experimental design

13
The purpose of using linked data in our approach is to formalize relationships between data hosted
through Esri Open Data or any other spatial repository, and publications hosted anywhere. The linked
data publishing pattern followed in this research generates linked data from static structured data in the
manner of Heath and Bizer (2011). This is achieved by taking static input data in the form of spatial and
non-spatial contents, publishing them as services and generating a triplestore to reference the URIs of data
services and associated publications. This is achieved with the aid of the tool OpenRefine and its RDF
extension. The stepwise procedure undertaken to achieve this is summarized as follows:
1. Data hosting: Spatial and non-spatial research data are published to a local server and shared via
ArcGIS Online as image or feature services, which are shared with the UCSB Open Data group
by a system administrator and are made publically referenceable through Open Data source URIs.
2. URIs: Identifiers for corresponding publications and dataset services referenced by Open Data
content are retrieved from open access document repositories or publisher pages.
3. Vocabularies: The OpenRefine with RDF extension generates a graph using the identifiers of
publications and research object relationships defined by Dublin Core predicates.
4. Reconciliation: The graph is referenced against Library of Congress Subject Headings to enrich
users’ ability to explore and discover thematically linked content.
5. Implementation: Publication-data relationships are serialized as triples that can be queried using
the SPARQL Protocol and RDF Query Language.
Because the data described in the previous section are hosted as web services, they are easily referenced
through their URIs. Researchers at UCSB can currently share their spatial and non-spatial research data
through the institutional instance of ArcGIS Online. Any content currently available through this platform
can be migrated into Open Data by changing system permissions. A small subset of data are currently
hosted for this research, but by hosting data directly on ArcGIS Online and by connecting additional
external resources associated with UCSB researchers to UCSB Open Data, we hope to expand content.
The URIs of research objects correspond to either a data layer or a publication. Datasets that
share a common base name are parts of collections and are indicated by URI container. Data creators have

14
only partial control over assignment of the URI resource domain name, which as a best practice, should
be self-descriptive and human readable (Heath and Bizer 2011).
When selecting a technology for RDF creation, it was important to consider the provided data
formats, mechanisms of access and desired output. While initial stages of this research tested the
Callimachus linked data application builder, a locally hosted triplestore was deemed to be inefficient and
limiting. Several other RDF converters and services were considered, but many of these tools perform
script-based extraction, transformation and loading from web pages. Semi-automatic RDF creation, rather
than script-based extraction for instance, is a technique better suited to our purposes.
The nature of the data and the questions asked about the data determine the choice of vocabulary.
Using predicates from existing vocabularies increases data interoperability and reuse. Other datasets and
applications that use shared vocabularies can also be more readily cross-linked without additional
processing, increasing their discoverability (Heath and Bizer 2011). The Dublin Core Metadata Initiative
(DCMI) vocabulary is widely used and is well maintained with dereferenceable URIs that point to a
retrieval protocol. These factors motivated the decision to use DCMI instead of specialized vocabularies,
which are typically less stable. DCMI metadata elements define general attributes such as title and
subject. Our data model does not rely on metadata standards, but rather on the two simple associative
relationships, isReferencedBy and references, defined in the Dublin Core ontology16 shown in Figure 3.
Figure 3—Generic Dublin Core Metadata Initiative (DCMI) data model
One of the motivations for producing linked data is to forge associations with other data sets,
which is a step achieved during the reconciliation process. URIs of the research objects can be interlinked
with Library of Congress authority files17 and even extended to link with other contextually relevant
ontologies, such as Wikipedia’s knowledge graph DBPedia18, by referencing the SPARQL endpoints.
These links enable exploration of other works associated with authors and datasets.
16 http://dublincore.org/documents/2012/06/14/dcmi-terms/?v=terms#
17 http://id.loc.gov/
18 http://mappings.dbpedia.org/server/ontology/classes/

15
Once the linked data model has been applied to the publications and dataset URIs, OpenRefine
generates an RDF skeleton. The interface allows users to preview the RDF schema and manually edit
nodes in the graph. Once the structure is formalized, it is possible to export the data to a variety of
formats, such as RDF/XML or Turtle, depending on the intended use, as shown in Figure 4.
Figure 4 — Reconciled OpenRefine template (above) and RDF skeleton (below)
We used OpenRefine with its RDF extension to implement our simple linked data model.
OpenRefine generates a static profile triplestore, which is an internally hosted RDFa document that
references the URIs assigned to the publication and research data. These static files can then be uploaded
to a web server, offering users a web-accessible interface that supports queries.
In OpenRefine, a class is a set of RDF resources that use the same templates. Classes such as
publications and data are defined as instances. A new Publications class template uses an RDFa
serialization, embedding RDF as triples in HTML documents and encoding the semantic properties and
relationships captured in Figure 5.
Figure 5 — RDF triples for datasets and publications exported in Turtle syntax
The triplestore can be queried using SPARQL. A SPARQL endpoint is a web-protocol to which
queries against a triplestore can be submitted (Powell 2014). User queries pertaining to datasets
referenced within a publication or publications that utilize a particular dataset can be formulated in this
way. General queries across all relationships as well as between specific publications or datasets can then
be generated. A SPARQL query for all publications that reference datasets can be formulated against the
triples, as shown in Figure 6.
Figure 6 — A generic SPARQL query against the triples
In this query, a user requests the attributes of data associated with publications, which are then
optionally filtered by matching author name and sorted by title. By formalizing the relationships between
subjects and objects through the use of DCMI prefixes during the data production phase, it is possible to
map the relationships between research publications and datasets. In this example, matching triples for all

16
publications referencing datasets produced by Dr. Anabel Ford are returned, sorted by title. Additional
queries could be constructed using any combination of predefined attributes and predicates.
4. Results
The research datasets tested in the model included a Maya Forest GIS layer featuring archaeological sites
on UCSB Open Data as the object and a published report from the researcher on the 2000 Field Season19
as the subject (Ford and Wernecke 2000), which are illustrated in Table 2. Queries for datasets associated
with a particular publication use the DCMI predicate references to point users to linked datasets hosted
through UCSB Open Data. Conversely, users can query for publications associated with datasets through
the predicate isReferencedBy, which points back to objects in their respective repositories using URIs.
Table 2—Example of a triple stored in the RDF framework
The two parameters defined within the OpenRefine template include a publication URI resource,
which is provided by the user, and a data URI resource, which in the case of the sample data comes from
Esri Open Data. The relationship between these entities is manually defined. The template references the
DCMI vocabulary and makes assignments to each resource based on the user asserted relationship.
OpenRefine with RDF extension offers a flexible template that can easily be extended to include
additional prefixes and connect the research objects to other collections.
Once linked data are generated from the research objects, publications and datasets are
discoverable from their URIs. Users can spatially browse for datasets through the UCSB Open Data
instance and discover linked datasets based on the associated attributes formalized in the data model.
Importantly, this process enables the spatial discovery of research publications associated with spatial
datasets, which are not traditionally conceptualized as objects with footprints. Retrieving datasets from
publications is also possible through the linked data model, as pointers to the hosted data can be exposed
during a search on an external repository.
19 http://escholarship.org/uc/item/4qr2x8p3

17
We have deliberately avoided developing a complex model of authorial relationships between
data and publications. With a data model for generating simple associative triples in place, scaling up the
number of resources referenced in the system from our current small set will be possible. User-testing to
ensure the data model is adequate, and achieving a critical mass of datasets and associated publications
will eventually result in a cross-disciplinary discovery resource.
5. Discussion and Conclusions
This article presents a first step in establishing a linked data discovery mechanism that prioritizes stability
and supports the discoverability and reusability of research data with spatial references, whether these are
in the data themselves or just metadata. It demonstrates how academic libraries can spatially enable the
discovery of research objects across disciplines and systems. Formalized relationships between
publications and researcher-generated data expose the interplay between researchers and the data that they
use or produce. Linking research data, hosted for example through Esri Open Data, to publications, such
as those accessible through the UCSB Alexandria Digital Research Library repository, adds value to both
sets of research objects. By creating links through the use of linked data predicates taken from Dublin
Core, library users are led from publications to data and back, leveraging the spatial search in Esri Open
Data on a much broader scale. Making an increasing amount of content compatible through a linked data
model will make more library holdings discoverable through a spatial search interface.
5.1. Limitations
Current institutional policies support research sharing through open-access licensing, yet incentives and
formal channels for sharing only currently exist for publications, not necessarily for associated datasets
(University of California Regents 2014). Therefore, in order to lower hurdles to participation, the system
described here opts to give researchers full control over what data they want to make available and how.
Another open issue is the long-term maintenance of such a system. The production of linked data
is currently a manual process undertaken using OpenRefine. Transitioning to a system that automatically
scrapes repositories and generates links may be desirable. The use of semi-automatic RDF creation in this

18
research enabled reconciliation of resources through a graphical user interface, yet this required manual
effort. The process could be expedited through the use of server-side tools like Apache Jena20 to automate
the workflow by running periodic scrapes and generating triplestores from URIs.
5.2. Next steps
Metadata for objects in ADRL recently became available as RDF triples, which are available through a
dedicated API. UCSB Library staff harvest metadata in ADRL from the MARC metadata in the library
catalog. Aligning this collection with the triplestores for research objects currently generated in
OpenRefine can increase the amount of campus resources accessible as linked data, expanding the
university’s knowledge graph. Linking these systems through common vocabularies could increase
awareness of research efforts across domains and increase discoverability of curated research objects
The ADRL efforts will also result in the eventual contribution of name records for all new
electronic thesis and dissertation authors to Library of Congress Name Authority Files, which are
referenced by libraries as a controlled vocabulary for bibliographic records. Name records will be
available through the Library of Congress Linked Data service as URIs. UCSB Open Data and the ADRL
content now available as linked data could readily reference these authority headings (Maali 2011).
Different linked data sets do not have to necessarily share a single schema, yet their structure
allows them to support cooperation without a need to coordinate. Adopting a linked data approach that
defines the relationship between objects regardless of their format, location, or metadata schema,
expands the scope of content discovery beyond that which any single system can offer. By extension, this
expands discovery beyond an individual campus to the broader research community.
6. Acknowledgements
We would like to thank the UCSB Library, UCSB Center for Spatial Studies, and Esri Inc. as well as
Anabel Ford and Ben Halpern for supporting this research project. The research reported here has been
partially supported by an anonymous private donor.
20 https://jena.apache.org/index.html

19
7. References
Addison, A., Moore, J., and Hudson-Vitale, C. (2015). Forging partnerships: Foundations of
geospatial data stewardship. Journal of Map and Geography Libraries 11(3): 359–375.
Athanasis, N., Kalabokidis, K., Vaitis, M., and Soulakellis, N. (2009). Towards a semantics-based
approach in the development of geographic portals. Computers and Geosciences 35(2): 301–308.
Avram, H. D. (2003). Machine-readable cataloging (MARC) program. Encyclopedia of library and
information science 3: 1712.
Baierer, K., Dröge, E., Trkulja, V., Petras, V. (2014). Linked Data Mapping Cultures: An
Evaluation of Metadata Usage and Distribution in a Linked Data Environment. In Proceedings of
the International Conference on Dublin Core and Metadata Applications.
Bechhofer, S., De Roure, D., Gamble, M., Goble, C., & Buchan, I. (2010). Research objects:
Towards exchange and reuse of digital knowledge. The Future of the Web for Collaborative
Science.
Brewer, G. D. (2015). The challenges of interdisciplinarity. Policy Sciences 32(4): 327–337.
Buckland, M. K. (1997). What is a document?. Journal of the American Society for Information
Science (1986-1998), 48(9), 804.
Cooley, S., Lafia, S., Medrano, A., Stephens, D., and Kuhn, W. (2015) Spatial Discovery Expert
Meeting Final Report. Center for Spatial Studies, University of California, Santa Barbara Library.
eScholarship: http://escholarship.org/uc/item/64p820kg.
Cragin, M., Palmer, C., Carlson, J., and Witt, M. (2010). Data sharing, small science and
institutional repositories. Philosophical Transactions of the Royal Society 4023–4038.
Durante, K., and Hardy, D. (2015). Discovery, management, and preservation of geospatial data
using hydra. Journal of Map and Geography Libraries 11(2): 123–154.
Durante, K., Weimer, K. H. and McGee, M. (2016) “Linked Open Data Modeling for Library
Cartographic Resources.” Presentation at the Annual Association of American Geographers
Conference, San Francisco, CA, March 29.
Ford, A., Wernecke, C. (2000). Assessing the Situation at El Pilar: Chronology, Survey,
Conservation, and Management Planning for the 21st Century. MesoAmerican Research Center.
UC Santa Barbara: MesoAmerican Research Center. eScholarship:
http://escholarship.org/uc/item/4qr2x8p3
Ford, A. (1995) Archaeological Sites Maya Forest GIS. WWW document,
http://discovery.ucsb.opendata.arcgis.com/datasets/1a3a1295bf2e4cafab64580182d15367_0
Freeston, M. (2004). The Alexandria Digital Library and the Alexandria Digital Earth prototype. In
Proceedings of the 2004 joint ACM/IEEE conference on Digital libraries—JCDL 2004, p. 410.
New York, New York, USA: ACM Press.

20
Golding, C. (2009). Integrating the disciplines: Successful interdisciplinary subjects. Centre for the
Study of Higher Education, University of Melbourne.
Goodchild, M. F. (2004). The Alexandria Digital Library Project. D-Lib Magazine, pp. 1–8.
Halpern, B. Lester, S. and Grorud-Colvert, K. (2009). PISCO: Partnership for Interdisciplinary
Studies of Coastal Oceans. Science of Marine Reserves: Meta-analysis: Global synthesis. KNB
Data Repository.
Hardy, D., and Durante, K. (2014). A Metadata Schema for Geospatial Resource Discovery Use
Cases. Code4lib Journal 25.
Harnad, S., Brody, T., Vallieres, F., Carr, L., Hitchcock, S., Gingras, Y., Oppenheim, C., Hajjem,
C., and Hilf, E. (March 2008). The Access/impact Problem and the Green and Gold Roads to Open
Access: An Update. Serials Review 34 (1): 36–40. doi:10.1016/j.serrev.2007.12.005.
Heath, T., and Bizer, C. (2011). Linked Data: Evolving the Web into a Global Data Space.
Synthesis Lectures on the Semantic Web: Theory and Technology, Morgan and Claypool.
Hey, T., Tansley, S., and Tolle, K., eds. (2009). The fourth paradigm: Data-intensive scientific
discovery. Redmond, WA: Microsoft Research.
Hu, Y., Janowicz, K., Prasad, S. and Gao, S. (2015-1), Metadata Topic Harmonization and
Semantic Search for Linked-Data-Driven Geoportals: A Case Study Using ArcGIS Online.
Transactions in GIS, 19: 398–416. doi: 10.1111/tgis.12151
Hu, Y., Janowicz, K., Prasad, S., and Gao, S. (2015-2). Enabling Semantic Search and Knowledge
Discovery for ArcGIS Online : A Linked-Data-Driven Approach. In AGILE, pp. 1–16.
Keßler C, Janowicz, K., and Kauppinen, T. (2012). Exploring the research field of GIScience with
linked data. In Xiao N, Kwan M-P, Goodchild M F, and Shekhar S (eds) Geographic Information
Science: Seventh International Conference, GIScience 2012, Columbus, OH. September 18–21,
2012, Proceedings. Berlin, Springer Lecture Notes in Computer Science 7478: 102–115.
Maali, F., Cyganiak, R., & Peristeras, V. (2011). Re-using Cool URIs: Entity reconciliation against
LOD hubs. CEUR Workshop Proceedings, 813.
MacMillan, D. (2014). Data Sharing and Discovery: What Librarians Need to Know. The Journal
of Academic Librarianship Vol. 40 (5): 541–549.
Mayernik, M. S. (2015). Research data and metadata curation as institutional issues. Journal of the
Association for Information Science and Technology,
Mobility, measured: America is no less socially mobile than it was a generation ago (2014,
February 1). The Economist. Retrieved from http://www.economist.com/news/united-
states/21595437-america-no-less-socially-mobile-it-was-generation-ago-mobility-measured.
Padilla, T. (2016). Humanities Data in the Library: Integrity, Form, Access. D-Lib Magazine, 22
(3/4), 1–12.

21
Patterson, T., Kelso, N. V., & North American Cartographic Information Society. (2009). Natural
earth. WWW document https://searchworks.stanford.edu/view/11047527
Powell, J. (2014). A Librarian's Guide to Graphs, Data and the Semantic Web. Chandos
Publishing. Oxford.
Regents of the University of California. “UC Open Access Policies Office of Scholarly
Communication.” Accessed January 12, 2016. http://osc.universityofcalifornia.edu/open-access-
policy.
Scaramozzino, J., White, R., Essic, J., Fullington, L. A., Mistry, H., Henley, A., and Olivares, M.
(2014). Map Room to Data and GIS Services: Five University Libraries Evolving to Meet Campus
Needs and Changing Technologies. Journal of Map and Geography Libraries 10 (1): 6–47.
Scheider, S., Degbelo, A., Kuhn, W., and Przibytzin, H. (2014). Content and context—How linked
spatio-temporal data enables novel information services for libraries. GIS.Science (4): 138–149.
Szukalski, B. (2015, June 17) "ArcGIS Online Demo: A Very Spatial Update." Spatial Discovery
Expert Meeting. Upham Hotel, Santa Barbara, California. Lecture.
Tsang, Daniel C. (2015). Academic Librarians & Open Access of Data: Challenges &
Opportunities in Research Data Management. UC Irvine: UC Irvine Libraries.
van Hooland, Seth and Verborgh, R. (2014). Linked Data for Libraries, Archives and Museums:
How to clean, link and publish your metadata. Facet.

22
8. Tables
Table 1—Personas, domains, and datasets of researchers currently discoverable through UCSB Open Data
Persona
Domain
Dataset
Dataset
location
Publication
Publication
location
Anabel
Ford!
archaeology!
Archaeological
Sites21!
UCSB Open
Data
Assessing
Situation El
Pilar22
UCSB
eScholarship
Benjamin
Halpern
ecology
Science of
Marine
Reserves:
Meta-analysis23
Knowledge
Network for
Biocomplexity
Biological
effects within
no-take
marine
reserves: a
global
synthesis24
UCSB UC-
eLinks
via WorldCat
Tom
Patterson
political
science
World
Boundaries of
Disputed
Areas25
EarthWorks
Natural
Earth26
SearchWorks
21 http://opendata.arcgis.com/datasets/1a3a1295bf2e4cafab64580182d15367_0
22 http://escholarship.org/uc/item/4qr2x8p3
23 https://knb.ecoinformatics.org/#view/doi:10.6085/AA/pisco_smr_synthesis.1.3
24 http://ucelinks.cdlib.org
25 https://earthworks.stanford.edu/catalog/stanford-tq310nc7616
26 https://searchworks.stanford.edu/view/11047527

23
Table 2—Example of a triple stored in the RDF framework
Resource
Subject
Predicate
Object
data
Archaeological Sites
Maya Forest GIS
isReferencedBy
Assessing the Situation
at El Pilar
publication
Assessing the Situation
at El Pilar
references
Archaeological Sites
Maya Forest GIS

24
9. List of table numbers and captions
Table 1—Personas, domains, and datasets of researchers currently discoverable through UCSB Open Data
Table 2—Example of a triple stored in the RDF framework
10. List of illustration numbers and captions
Figure 1—Project vision for data discovery and publication integration across domains

25
Figure 2—UCSB’s Open Data instance27 leverages ArcGIS Online
Figure 3—Generic Dublin Core Metadata Initiative (DCMI) data model
27 http://discovery.ucsb.opendata.arcgis.com/

26
Figure 4 — Reconciled OpenRefine template (above) and RDF skeleton (below)

27
Figure 5 — RDF triples for datasets and publications exported in Turtle syntax
Figure 6 — A generic SPARQL query against the triples