Enhancing Accessibility and Interoperability of Mobility Data: MobilityDCAT-AP, a Metadata Specification for Mobility Data Portals

How to make different data assets from different actors in the mobility system dis-coverable and accessible? This is the role of internet portals for mobility data which have been developed all over the world in recent years. These portals often have specific spatial or thematic coverage, e.g., providing public-transport timetable data for a specific region. In addition, legal obligations, such as those established through the European Union's National Access Points (NAPs) [1], mandate the creation and population of such portals.

This work aims at supporting potential data users during the discovery and assessment of data offerings in NAPs and other mobility-related data portals. Metadata can be particularly helpful in understanding the features and options of corresponding data, especially in complex and ever-changing ITS environments [2]. Thus, this work enables to describe data offerings via metadata, answering questions such as what data are offered, by whom they are published, what access conditions apply, etc. More specifically, the work structures and harmonises such descriptions in the form of a formal metadata specification called mobilityDCAT-AP. This specification provides precise and unambiguous metadata designations for any mobility related data offerings, facilitating harmonised, platform-independent metadata descriptions both in human-readable and machine-readable formats.

An important driver for such metadata specification is found in the context of interoperability. The mobility data ecosystem consists of different elements that complement and relate to each other. For example, a single data set may be bound to a single use case or a single region. To get a complete picture, multiple data sources from multiple portals may need to be combined. However, handling multiple data sources requires that the corresponding metadata are interoperable in terms of semantics and syntax. Interoperability depends on domain-specific metadata specifications or standards [3]. They aim to provide an agreed and uniform way of describing metadata, regardless of the metadata origin or user. They precise formal structures, vocabularies and agreements on their use. Another driver is the need for efficient ways of storing and sharing metadata, giving a key role to Resource Description Framework (RDF), which is also a prerequisite for Linked Data and Semantic Web concepts [4]. As a result, many of the established metadata standards and specifications are based on RDF.

This work has been carried out in the context of NAPCORE, an EU-cofunded Programme Support Action under the GRANT AGREEMENT No MOVE/B4/SUB/2020-123/SI2.852232 [5].

One well-known framework for defining metadata and specifications is the Data Catalog Vocabulary (DCAT), as an RDF vocabulary designed to describe data catalogues [6]. DCAT Application Profile (DCAT-AP) specifies DCAT as a basic profile for data portals in Europe to facilitate the aggregation, exchange, search and auto-mated processing of metadata [7]. DCAT-AP scope is cross-border and cross-domain, and thus further specified in different, domain-specific extensions like GeoDCAT-AP for spatial data [8] and StatDCAT-AP for statistical datasets [9]. In the field of mobility, some extensions such as TransportDCAT-AP in the domain of public transport, as created in the OASIS project [10], and spsDCAT-AP in the domain of Smart Parking Systems are based on DCAT-AP [11].

The Shift2Rail IP4 SPRINT project provided an automated solution for ingesting and harmonising metadata from selected NAPs to enable cross-border multimodal journey planning [12]. Similar work was carried out in the LOD-RoadTran18 project [13], which includes a specific metadata model to represent road-traffic information datasets. Both projects again rely on the DCAT-AP vocabulary.

Besides such DCAT-AP-based approaches, there are some proprietary metadata specifications, i.e., approaches based on proprietary metadata models and ontologies. The Coordinated Metadata Catalogue of the EU EIP project was a first common blueprint for metadata structures for NAPs in Europe [14]. It defines a common minimum set of metadata, including descriptions, formatting constraints and obligation levels. However, these definitions are only available in a proprietary, human-readable format.

This review identifies some previous approaches to metadata specification and harmonisation. However, many of these have not been developed for the generic domain of mobility. Moreover, some seem to be at an experimental level, not ready for community-wide and long-term application. Finally, some did not fully address interoperability and conformance with semantic technologies via RDF. Therefore, the presented work is a first approach towards a domain-wide, application-ready and RDF-compliant metadata specification in the domain of mobility data. It adds the mobility domain to the DCAT-AP extension family, with an own extension called mobilityDCAT-AP.

As a starting point, the NAPCORE Metadata Working Group [5] defined a roadmap for the design, implementation and publication of the envisaged metadata specification. A first action was a detailed requirement analysis involving experts and mobility-data stakeholders [15]. Based on this, a conceptual model was developed, which translates the before-formulated requirements into a technology-independent data structure. We focused on essential information about a mobility data offering on a data portal, and analysed common practice in NAPs and other mobility portals, harmonisation exercises, and feedback from experts. Figure 1 shows an aggregated view of the essential aspects identified.

Bild vergrößern

Such essential aspects were then disaggregated in much more detail, also considering mobility-related characteristics. For example, the aspect “Content info” was specified by the transport mode, the part of the transport network and other details covered by the data offering. The final conceptual model was developed iteratively in a tabular format, listing essential metadata elements including definitions, obligation levels, and usage notes.

The elements of the conceptual model were then mapped to corresponding RDF elements. RDF data representations are based on a data model architecture that includes classes and properties. As mobilityDCAT-AP is planned as an extension to DCAT-AP [7], the first step was to check if and how existing elements from the DCAT-AP vocabulary correspond to the elements of the conceptual model. In fact, many existing elements could be reused, although some definitions and usage notes had to be rewritten, and some obligation levels had to be changed. Further, additional classes and properties from other vocabularies were incorporated. For example, the quality description of a data offer, being an essential aspect to be provided via metadata, was considered by reusing the established Data Quality Vocabulary [16], which is another RDF vocabulary.

In cases where neither DCAT-AP nor any other known RDF vocabulary provided applicable elements, mobilityDCAT-AP added its own vocabulary elements. These additions aim at capturing some specific characteristics and features of mobility data. This is done by creating a separate namespace “mobilitydcatap:” and individual class/property declarations. Among other things, a property “mobilitydcatap:transportMode” has been introduced this way. MobilityDCAT-AP has also re-moved some optional properties from DCAT-AP. The reason for this is the rather broad scope of DCAT-AP, and the aim to keep the vocabulary size of mobilityDCAT-AP as compact as possible.

The final data model, comprising all RDF elements representing the conceptual model, was visualized as a Unified Modelling Language (UML) diagram. Figure 2 shows central classes from this diagram, and selected properties, as embedded within such classes.

Bild vergrößern

Accordingly, four central classes represent a hierarchical concept when describing metadata via mobilityDCAT-AP. Firstly, the Catalogue as such is described, i.e., being the metadata register in a data portal. Secondly, there is the Catalogue Record, which describes the metadata entry, including its publication date. Thirdly, the Dataset is described. In fact, most metadata elements are covered here, including the content theme; the spatial and temporal context; quality information and others. Finally, the distribution describes a technical and organisational way to access the Dataset. In addition to the data format (e.g., a machine-readable syntax standard), the licensing terms are described here. Some additional classes (not shown above) act to support these general classes.

Some of the vocabulary elements are restricted in such a way that the possible expressions of the vocabulary are bound to predefined value lists. Such value lists are stored under Controlled Vocabularies. For example, the Controlled Vocabulary of the above-mentioned property “mobilitydcatap:transportMode” contains the values “car, truck, bike, pedestrian, etc.”. Such predefinitions aim at the unambiguous use of frequently used terms. In addition, Controlled Vocabularies are also expressed as RDF vocabularies and allow interoperable processing. For example, a Controlled Vocabulary could be provided in multi-lingual versions.

There is online documentation of the mobilityDCAT-AP specification [17], including a human-readable vocabulary description; a set of usage notes; a set of machine-readable serialisation formats to allow IT systems process the vocabulary; and schemas for automated conformance validation. In addition, user support is provided via a dedicated GitHub repository where users can interact with the editors, for example by raising questions and issues.

Finally, there is an Accompanying Guideline that serves as a practical orientation for users of mobilityDCAT-AP. This includes additional explanations and examples for specific vocabulary elements; recommendations for metadata handling and exposure on individual IT systems; and advice on metadata quality and validation processes.