In many research projects, much more data are created than made publicly available. Keeping research data deliberately closed or publishing only selected subsections of the gathered data are unfortunately common practices in academia. Fortunately, such problems have been getting more and more attention in the past years. However, another issue that is still often overlooked concerns research data that are generated as part of a research project but that are generally not considered part of the primary research data. One example for such neglected research data are terminologies such as controlled vocabularies that are used to describe or classify primary research data. In this paper we will outline the process that is used by the Terminology Service of the German Federation for Biological Data (GFBio) to prepare and process terminologies so that they can be included in the GFBio Terminology Service where they are made available to researchers within and outside the original research project. We will also show how making such supplementary research data publicly available will benefit the researchers who share them as well as the scientific community as a whole.
In recent years, primary research data have been getting more attention as part of the
publication process. Funding agencies such as the German Research Foundation (DFG3)
and publishers are pushing scientists to publish the underlying research data along with
the corresponding papers, or at least upload them to research data repositories. The
DFG-funded project GFBio4 (German Federation for Biological Data) is creating a
dedicated repository for various kinds of biological research data and is developing
supplementary tools for discovery and reuse of these data [
Terminologies that are used to describe or classify primary research data can therefore be considered as supplementary7 research data, data that is not the primary focus of a research project, but vital to the accumulation of the primary research data. 2
2.1
In the context of the GFBio Terminology Service and this paper, a terminology is the overarching name for any set of fixed denotations that are used to describe something with the goal to reduce ambiguity and facilitate comparability. A terminology can range from a simple Controlled Vocabulary (a simple list of terms) to a complex Ontology (formal definitions of terms and their relations semantically expressed in a machine readable way). Terminologies can include translations and synonyms or aliases for individual terms. 2.2
The German Federation for Biological Data (GFBio) is a national data infrastructure
to store and facilitate access to biological and environmental research data. It offers
services and resources to researchers for the archiving and publication of their research
data as well as an open access portal to provide access to the data stored in the various
data centers. The Terminology Service8 (TS) of GFBio provides access to various
terminologies for research data through one unified API [
Different systems providing a comparable terminology service exist, the most widely
used being Bioportal [
If researchers want to have their terminology included in the GFBio Terminology Service, they need to contact the TS team, either directly or through the GFBio Submission Page13. To make a terminology fit the requirements for import in the Terminology Service, several processing steps might be required. These steps are done in close cooperation between the TS team and the scientist(s) providing the terminologies. The steps strongly vary between the individual terminologies, their type and complexity, and the additional work already provided by the involved scientists. The simplest case is when a dedicated list of terms is available as part of the supplementary research data, ideally with definitions and connections between the terms. In cases where no dedicated list of terms or formal documentation is available, the terms are extracted from the primary research data. This can range from simply exporting individual columns or tables from the set of the primary research data to doing complex parsing operations on the data to filter out the desired terminologies. The software used to do these extractions depends on the original data, e.g. when the terminology is included in the form of geographic data files, a common GIS software is used to extract it. The goal of the extraction process is to end up with a tabular file of the individual terms and their corresponding information, like hierarchies, if they can be extracted as well. Once the extraction is done, the scientists are asked to review the information for the completeness and correctness and provide any missing information that were not part of the original research data, such as definitions, translations or hierarchical structures in cases where they could not be extracted. The next step of the terminology processing is the data refinement and cleanup, which again is done in close contact with the contributing scientist(s). The refinement is usually done using OpenRefine14, to catch errors like spelling mistakes in the term names, resulting in two very similar but not identical terms. Different additional tools are sometimes also 13 https://www.gfbio.org/data/submit/generic; This is the same page as for the general GFBio data submission. 14 www.openrefine.org used to check for logical errors in the structure or other errors that cannot be checked using OpenRefine.
Each term of the terminology will get an individual URI which makes them addressable as a resource in the Semantic Web context. To avoid creating additional URIs for the same concepts, similar terminologies are searched for and if available, their terms are compared to the terms of the current terminology. In cases where terms are identical, the already existing URI is used. If terms are comparable but not identical to terms from other terminologies, then the relation between the terms is recorded by using properties such as skos:broader or skos:related. There are two options for contributing scientists if new URIs for the terms are assigned. The terms can either get the GFBio TS prefix15, or they can provide their own prefix. The URIs with the TS prefix are resolvable and provide both human and machine readable formats depending on who is resolving the link. Custom URI prefixes on the other hand can help the branding of research projects, but the researchers are responsible for resolving the terms if they wish to have this highly recommended feature. In the end, the metadata of the terminology itself are formalized and the terminology is exported. Depending on its complexity this is usually SKOS, OWL or another RDF-based format which can then be imported into the GFBio Terminology Service. The export is done by creating a template in which the individual terms can be imported and using the OpenRefine templating engine to generate the final RDF file. After a final check and validation, the file is then imported into the GFBio TS, where the terms are then accessible via the TS API. When several scientists from the original research project wish to collaboratively and simultaneously work on reviewing and extending the terminology during the different feedback steps mentioned above, the TS team can provide dedicated tools. 4
There are several advantages that come with having research terminologies accessible.
The foremost gain is that the primary research data itself becomes more understandable
and reusable when the definitions and underlying hierarchies of the terms used to
express it, are available as well. This is the primary use case of supplementary research
data. These advantages can be further extended if the primary research data are served
through a semantic aware search or portal, as this will allow for queries that also include
synonyms or higher hierarchical terms, as demonstrated in [
The GFBio Terminology Service is an important resource both for scientists who wish to share their terminologies that are used to describe and classify research data and for researchers who wish to apply existing terminologies and classifications to their own research data to improve their integrability. With reasonable additional effort the terminologies can be processed to be included in the TS and both the scientists who created the terminologies and the scientific community as a whole can benefit from this otherwise neglected research data.