Ontologies land at the airports

Managing complex aeronautical data is a challenge for any organisation

Managing complex aeronautical data can be a significant challenge for any organisation, whether it is a government agency, an airport, an airline, an aircraft manufacturer or a service provider.

To win that challenge, models of greater or lesser complexity are normally developed in order to characterise, organise, manage, use and archive the data generated inside the organisation.

In an interconnected world, however, a company is only one of the many actors that produce data (data providers), and each of those actors adopts a data model that differs in quality from the others. That is where integration problems come from. Over the last decade those problems have pushed data producers and consumers towards standardised data models: it is now widely recognised that tackling the interoperability of enterprise data with data coming from the outside world means enriching company knowledge and increasing competitiveness.

For some years now Tim Berners-Lee has identified the technological solutions for evolving the whole web towards a new dimension, one able to release all the potential contained in data and to solve interoperability and privacy problems at their root. In that landscape, data models based on ontologies play a relevant role — in the aeronautical world too.

Ontology-based models have emerged in recent years from the confluence of research on artificial intelligence and on the Semantic Web, in order to integrate heterogeneous and distributed data and to help resolve inconsistencies and errors. The new Solid project, strongly advocated by Tim Berners-Lee, likewise uses linked data and ontologies as the foundations of what will very probably represent the next significant step in the evolution of the Internet and of the business connected to it.

Several use cases for these technologies already exist in data management for the aviation and airport sector.

An opportunity for the civil aviation market

The volume of data of interest to civil aviation, generated by industry and by government sources, has grown steadily over the last ten years, and an ever-increasing number of actors has begun publishing information through various digital distribution methods. The landscape of Open Data, freely available on the Internet, has also undergone an incredible acceleration, with a decisive increase in the quantity and quality of the data available.

This trend is a great advantage for those who use the information to improve air transport systems, to obtain cost savings, to improve safety and to analyse bottlenecks. But alongside the advantages, an increase in the complexity of data management has to be reckoned with.

Aeronautical data providers employ different data models, and those models differ along several dimensions:

• Data encoding format. Providers may use different formats to encode their data. The airline ID field, for example, may be stored in a variety of ways: as a three-character field (International Air Transport Association, IATA), as a four-letter code (International Civil Aviation Organization, ICAO), or as a key referring to a table in the carrier’s database.
• Data field naming. Without standardisation, the field names assigned to values can be misleading and can lead to confusion. To describe an “aircraft departure time”, one provider may use the name dptTime while another uses DT. Reconciling these differences between models may be far from trivial.
• Data semantics. Even when two data fields carry identical names, it is not guaranteed that the data represents the same information. The actual meaning of “aircraft departure time” can differ greatly between two providers: it may refer to the departure time from the gate or from the runway. That difference is impossible to determine when the field names in the two data models are identical, without more specific contextual information.
• Spatial and temporal resolution. A datum may be recorded at different temporal frequencies (for example once a minute versus once a second) and aggregated over different time periods (for example every hour, every year) or spatial regions (for example across airspace, geographic regions or political entities). Without conveying that knowledge of spatial or temporal extent, the data cannot be interpreted correctly or combined with other data.
• Conventions and units of measurement. Units are often omitted and left implicit in data storage schemas. This can lead to problems when different units are used across different systems.

In fact, every data provider has a view of the world that makes sense for its own business, and that perspective drives the way its data is structured, stored and transmitted. But that view of the world may prove difficult for the consumer of the data to interpret. An aircraft-engine manufacturer, for example, may structure and transmit engine data as a series of separate tables, each describing one aspect of the data: technical specifications, performance characteristics, prices, maintenance and so on. It is common for each table to include rows containing the data values for the different engine models. Yet whoever wants to use that data may be interested in one specific aircraft only. In a case like that, retrieving and assembling the desired data is complicated.

With all these potential differences, working with data from several providers can be extremely demanding. To answer that problem, the aviation industry has developed a number of standard models to facilitate data exchange and promote interoperability.

Using standards simplifies the problem but does not solve it. Very often, in fact, data arrives from providers belonging to different markets that use different standards, or from sectors where standardisation does not exist. Think, for example, of the usefulness of the information held on Wikipedia, or of the information supplied by Airbnb, which for obvious reasons does not conform to aeronautical standards but can be very valuable for travellers.

An innovative approach consists in using ontologies: highly composable tools that allow data to be given a formal meaning (semantics).

Figure: ontologies support all aeronautical data and contribute to an Enterprise Knowledge Graph.

Ontologies can support all aeronautical data and contribute to building an Enterprise Knowledge Graph — a reservoir in which all the data coming from various providers is linked through shared meanings (Linked Data).

The Enterprise Knowledge Graph is in turn the beating heart of a Smart Data Management Platform that supports the integration of all data sources organically. The Enterprise Knowledge Graph and the Smart Data Management Platform enable the Data Driven Enterprise paradigm: a company in which data and its analysis guide the strategic and tactical choices made by management. All the leading successful companies of our time can be described as data driven. Think of how Facebook, Google, Airbnb, Amazon, Alibaba and Tesla use data to create or modify their products.

In 2015 the Semantic Web for Air Transportation (SWAT) interest group was set up, bringing together 25 international participants — among them government agencies (FAA, NASA, EUROCONTROL, the Department of Transportation (DOT) / Volpe, the Department of Defense (DOD), Air Services Australia), standards bodies (OGC — Open Geospatial Consortium, ICAO), university laboratories (MIT Lincoln Labs) and industry (Mosaic ATM, Frequentis, Securboration, Alion, Booz-Allen-Hamilton). The group’s purpose was to exchange information on ongoing projects, use cases and shareable resources such as ontologies, development tools and implementation strategies.

Some of the applications that use ontologies today are:

• NASA’s air traffic management data integration. NASA has developed a prototype data-integration system demonstrating how ontologies can be used to integrate, query and search across heterogeneous air traffic sources — for example air traffic management (ATM) data, FAA data and data supplied by NOAA (the National Oceanic and Atmospheric Administration) and other providers. In that prototype a common ontology is used to link several types of data model and to enable queries across combined data sets.
• FAA Enterprise Information Management. The FAA has developed an ontology to standardise the terminology used within its own documents, in order to make information easier to find and, ultimately, to improve decision-making processes.
• SWIM controlled vocabulary. The FAA has also developed a controlled vocabulary based on SKOS that provides a single source of terms and definitions for the documentation relating to SWIM. SKOS (Simple Knowledge Organization System) is a W3C-standardised ontology used to represent controlled vocabularies, thesauri and taxonomies. SKOS can be used wherever the use of vocabulary needs to be regulated, including in multilingual contexts.

Use cases

This section describes a set of use cases for ontologies in the aviation sector.

Integration between airport services

Non-aviation business is today one of the largest sources of revenue for whoever manages an airport. The term Non Aviation Business identifies a broad and varied range of commercial services offered to the passengers, operators and visitors of airports — whether managed directly or sub-conceded to third parties — as well as real-estate activity.

Ontologies and the integration with open data enabled by Linked Data make it possible to develop non-aviation services personalised to the needs of individual passengers: displaying the weather forecast for the passenger’s destination, integration with tourist facilities or professional services, personalised and targeted advertising and communication. But also raising users’ awareness of rules, customs and habits at the departure and destination sites.

Monetising knowledge

If company knowledge is formalised through ontologies, it can be weighed and transferred objectively. That enables a company’s information patrimony to be entered as a valued asset on the balance sheet. In other words, every single piece of information — identifiable, traceable and licensable thanks to the Knowledge Graph — can be associated with an economic value based on objective market valuations.

For example, the spending propensity of a user arriving from a certain destination and heading to another is information easily derived by cross-referencing travel data with the spending carried out in the non-aviation area. Such information, computable anonymously without infringing any privacy right, can have significant value for many actors in the tourism sector, as well as for the planning and organisation of commercial spaces and areas. This value can be entered on the balance sheet because the information it refers to is all identifiable and traceable thanks to the ontologies that give meaning to the data held in the Knowledge Graph, and that information has a negotiable market value.

Travel planning

The shortest route to a destination is not always the best one. The data held in an Enterprise Knowledge Graph can enable applications such as the Lodmap2D product of LinkedData.Center, which is able to suggest a route between two places taking the interests of the individual into account. Even inside the same airport, it is possible to suggest a route that takes account of an individual passenger’s habits and available time — all of it to offer the best “customer experience”, combining the passage through airport areas with a visit to commercial spaces that are certainly of interest to that individual.

Interoperability of software systems

Ontologies can be used to facilitate communication and coordination between software systems with different programming interfaces. An ontology can serve as a lingua franca for translation between different systems. To that end LinkedData.Center has developed a specific product, Onto40, for industry 4.0, which can be used to integrate the various automated systems present at an airport intelligently. The data collected in the airport setting can then be made easily available to other airport operators whose systems are not necessarily integrated.

Integrating information from multiple sources

Semantic data integration is a key use case for ontologies. An ontology can serve as a scaffold onto which to overlay information from heterogeneous sources, forming a single integrated data source with coherent spatial, temporal and conceptual bases. This is, for example, the approach followed by NASA’s ATM data-integration project. The most precious characteristic of Linked Data is that it allows the complete traceability of the provenance of every single datum, and therefore an assessment of its quality, risks and value.

Searching and analysing data

The interlinked data structure of an Enterprise Knowledge Graph lends itself to building intelligent search engines that simplify the retrieval of data and the development of analytical applications — including, and especially, those based on A.I. and machine-learning algorithms.

Standardising terminology

A common use of an ontology is the standardisation of terminology, as happens in the SWIM project. Another example comes from the standardisation of the description of aircraft make, model and series.

Automatic inference

Using the rules and formal properties associated with ontologies, existing data can be verified and new data can be generated from existing data through logical rules borrowed from A.I.

Publishing data

Ontologies play a key role in Linked Open Data (LOD), the standard way of publishing data on the web. LOD is also the best practice recommended by the Agency for Digital Italy for publishing data on the web. Linked Data enables data marketing actions, which are increasingly important for increasing the visibility of one’s activities on the web.

Information extraction

A final use case for ontologies is the extraction of semantic information from unstructured text documents. This is achieved by matching the terms and relations analysed from natural-language text against terms and relations in a controlled vocabulary.