GSTAR IV: Return of the GeoJSON
Following on from my Days of Archaeology in 2013, 2014 and 2015 (and for the last time), the bulk of my Day of Archaeology this year focussed on my doctoral research, writing up my thesis on Geosemantic Technologies for Archaeological Research (GSTAR). It’s been a busy three years but the project is nearing completion and will hopefully inform heritage management and research strategy over the coming years.
The aim of the project was to show how geosemantic technologies can be used to provide a framework for working with heritage data in a range of research contexts. To this end, I have built a demonstrator application which is based around a map (obvs!) for the Stonehenge landscape and which draws data from Historic Environment Records, museums and project archives, allowing users to ask questions across these diverse resources taking advantage of the semantic goodness of Linked Geospatial Data, thesauri and ontologies. Geosemantic ‘glue’ was used to integrate horizontally between resources (such as monuments and artefacts found within or nearby) and vertically (ie between excavation records and monument/event HER records and museum collection records).
The ontologies used were the CIDOC CRM, CRM-EH and GeoSPARQL which allow the concepts used by the various sources to be aligned whilst the terminology provided by the thesauri (published using SKOS) allow for the various terms used to document these concepts to be related. In other words, the semantic tools allow for the different sources to be made interoperable and queryable with the results displayed and interacted with on a map.
Moving forward, the approach taken and successfully demonstrated could be scaled up to act as the basis for the next generation of heritage information portals; think of the Heritage Gateway but with some additional bells and whistles:
- the ability to undertake proper geospatial queries and analysis, even where there is no GIS data
- spatial queries mediated using geospatial semantics, to get away from purely Cartesian views of space dependent on geometry and the problems that entails for historic information
- complex querying across all of the participating providers, with differences in terminology ironed out
The demonstrator application is built using a range of standard web and geospatial technologies. Currently, the accessioning process for data is largely manual, built using the STELLAR Toolkit to process outputs from MODES and HBSMR, two major software packages used in museums and HERs respectively. A next step would be to automate this, which would be fairly straightforward from a technological if not a political perspective. If an automated pipeline could be implemented across all the HBSMR and MODES using institutions and organisations, this would cover an enormous amount of heritage information and, combined with a map based portal and live feeds to desktop GIS, would greatly improve the way in which we undertake all kinds of research activities, both in academic and commercial contexts.
Information from site archives was a little tricksier, as one might expect; such data does not typically get archived in a readily useable fashion unlike information found within the structured systems used for managing Historic Environment Record data or museums collections. However, with ongoing work relating to the digital capture and sharing of fieldwork information through OASIS, HERALD and the broader Heritage Information Access Strategy (HIAS), we are undoubtedly moving towards a time when this becomes not just possible but the norm. When this happens (and note I say when not if!), we can start to extend Linked Data principles more fully to our information resources, so monument records can be directly built up from linked fieldwork records, museum collection artefact records can be layered on top of linked excavation finds records and, on top of all this, our Research Agendas and Frameworks can be truly data driven, dynamic resources drawing directly on this web of Linked Data, informing and informed by ongoing research and our shared knowledge of the past, across all of our information resources.
The use of such geosemantic ‘glue’ allows for a much more intelligent approach to finding and working with geospatial information from heterogenous sources split across numerous providers. Take the following query for example:
Show me all the Bronze Age mounds where dolerite has been found during excavations and carved chalk balls were discovered nearby.
Using the HeritageData Periods thesaurus, it is possible to mediate different uses of language across sources to describe time-spans relating to the Bronze Age, using broader, narrower and/or related terms. We can use the FISH Event Types Thesaurus to find event records relating to interventions (including excavations) and draw on the project archives for these to check for finds of dolerite, potentially using geological ontologies such as GEON to mediate identifications of rock types. Using the FISH Object Types Thesaurus, it is possible to do the same for chalk balls or any other artefact type. Geospatial information may well not exist for these objects as recorded in museums collections, most likely not in the form of British National Grid coordinates at least, particularly where they were discovered in antiquity. But we do often have some basic spatial information such as an associated location (eg Stonehenge), parish (eg Amesbury) or named place (eg Stonehenge Road); in such cases we can use the Ordnance Survey Linked Data plus some of the spatial relationships defined by the Simple Features specification (used by the GeoSPARQL ontology) to perform a spatial query using these index terms via a bit of geosemantic magic. Moving forward, we can align our research questions with such resources and queries so, for example, if the dating of carved chalk balls (typically thought of as of Neolithic origin) were to change, we can use the same approach to identify contexts where such changes would have a knock on effect or where our broader understanding of deposits, sites and complexes may also need to be updated or where new research questions arise. So this may be the end of the GSTAR project, but it’s only just the beginning for the use of such approaches within the heritage sector.
Many thanks again to everyone who has helped, contributed and otherwise supported this research project along the way, particularly:
- Doug Tudhope, Alex Lohfink, Mark Ware & Ceri Binding (University of South Wales)
- Chris Brayne (Wessex Archaeology)
- David Dawson (Wiltshire Museum)
- Adrian Green (Salisbury Museum)
- Keith May (Historic England)
- Melanie Pomeroy–Kellinger (Wiltshire Council)