It’s about time: historical periodization and Linked Ancient World Data

Most of us who have taught students about the material remains of the ancient world would agree that time in the past is harder to conceptualize than space in the past. This is especially true of the conventional temporal divisions -- "periods" -- we use to group and categorize the art, archaeological remains, and historical events of the ancient world. When asked to mark ancient places on a map, students excel. When asked to place objects within a historical, archaeological, or art-historical period and to assign a corresponding date range, however, the same students sometimes seem to be picking terms and dates at random. We can see ancient places inscribed in present spaces, and we can imagine ourselves in them. But ancient time undergoes a curious flattening process, in which developments that took place over centuries appear to have happened all at once in a great jumble. Introduce into this confusion the idea that "periods" are essentially arbitrary conventions about which scholars disagree, and many students will decide that the periodization of the ancient world is simply impossible to learn.⬈#p1

Similar problems apply to the role of temporal divisions in the Linked Ancient World Data environment. Conceptual entities that we can associate with concrete physical manifestations -- people, places, texts -- have provided the foundations for a growing Linked Ancient World Data infrastructure in which it is increasingly easy to discover semantically-related information. Authoritative spatial gazetteers like the Pleiades platform have not only made it possible to extract maps from texts by matching place names to geographic coordinates, they have also provided a conceptual connection that initiatives such as the Pelagios Project (Simon, Barker and Isaksen 2012) have been able to use to bring together material associated with the same site in different databases (for example, material associated with ancient Athens). The use of Uniform Resource Identifiers (URIs) for people, places, and things sidesteps the barriers to discovery created by the use of different metadata schemata, different vocabularies, or different languages. At the same time, however, it requires general agreement about the identity of the underlying entity: Athens, Athenae, Athen, Afen, Athina have to refer to the same socio-political unit located at more or less the same geographic coordinates. ⬈#p2

Periods have been modeled as abstract components of formal ontologies (e.g. the CIDOC-CRM: http://www.cidoc-crm.org/crm-concepts/#_Toc224984167; Doerr, Kritsotaki and Stead 2005; Doerr, Kritsotaki and Stead 2010), and directory or gazetteer-based approaches have been attempted on several occasions -- for example, in the Electronic Cultural Atlas Initiative (Petras et al. 2006) and the Common Eras project (Isaksen et al. 2009). But periods have proven to work poorly with Linked Data principles, which require well-defined entities for linking. As concepts of convenience, often invented by a single scholar in a seminal work but then modified and debated over long periods of time, deliberately left vague in temporal terms in order to protect their usefulness for relative chronologies, individual periods have no objective existence. Yet they are deeply entangled with both physical space and absolute time, the latter measured by concrete calendar systems and ultimately by the physical rotation of the earth and its movement around the sun. Particular period terms are associated with particular absolute dates in particular places; in turn, changes in the identity of a place at a particular geographic location can only be described in terms of the passage of time, often in the form of historical periodization.⬈#p3

The creation of an agreed-upon controlled vocabulary for periods therefore requires either highly local definitions or definitions that are so broad and vague as to be all-encompassing. Both types exist: local vocabularies in use in the UK heritage community, for example, define the "Iron Age" as a period (presumably only in Britain) lasting from 800 BC to AD 43, the year of the Roman conquest (e.g. http://www.fish-forum.info/i_apl_e.htm), while the Getty Thesaurus of Art & Architecture, which seeks to offer more universal concepts, describes a global "Iron Age" in terms of the three-age system and notes generally that it has different dates in different places. It also offers a facet for the European Iron Age, which has several stylistic-chronological subgroups: here, the only value associated with Britain is the La Tène period/culture, which is placed in space in "Northern Europe and the British isles" and in time from "the mid-fifth century BCE" to the "mid-first century BCE". Though this definition and that of the UK heritage community overlap partially in time, space and concept, they employ different terms; conversely, a single period term such as "Iron Age" can change dramatically in meaning according to where it is used, or by whom. For Linked Data purposes, then, this is not a simple question of using URIs from a shared gazetteer to align different terms for the same period concept. A solution is needed that recognizes the fact that different scholars can conceptualize the same time range or cultural phenomenon quite differently, while others use the same terms to describe time ranges or cultural phenomena that differ to varying degrees.⬈#p4

I first encountered this problem during the development of GeoDia, an interactive spatial timeline of ancient Mediterranean history and archaeology (Rabinowitz 2013). The original design was focused on archaeological and art-historical periodization, as a way to explain visually to students how ancient remains were grouped together (Figure 1).⬈#p5

We began with the idea that we would have general, Getty-AAT-style periods that would then be attached to individual archaeological sites. But major problems with this approach soon began to emerge. Using a generic periodization meant, in some cases, that sites would be represented with periods that were not used to describe them in the literature. In other cases, it meant a conflict between the date-range of a period to be represented in the timeline and the history of a particular site. Pompeii is a good example: though Pompeii exists during the more general "Flavian period", that site ends abruptly in AD 79, in the middle of that period, with the eruption of Vesuvius. The idiosyncrasies of particular regions and sites therefore led us to a data-model in which the basic unit was the "site period", in most cases a local manifestation, with a site-specific date range, of a more general period concept (Figure 2).⬈#p6

Because GeoDia, as a synchronized mashup of a map and a timeline, relies on places with geographic coordinates, it was relatively easy to join the Linked Ancient World Data ecosystem, at least peripherally, by attaching Pleiades URIs to its sites. But it quickly became apparent that allowing links back in was impossible, because the interface does not display "places": it displays a dynamic combination of location and "site periods" drawn from separate tables in the underlying database. We could consume URIs, but we couldn't produce them. Furthermore, while Pleiades URIs could be pulled in on a semi-automatic basis, it was impossible to connect GeoDia periods or site periods with controlled vocabularies. It was therefore impossible to take advantage of the possibilities for automated data integration offered by Linked Data approaches -- for example, the association of Flickr photos with ancient places in Pleiades through the use of URI-based machine tags. Frustration with this situation led me to join what I soon realized was a growing chorus of voices calling for the better integration of time in spatial approaches to history (Gregory 2007, Berman 2011, Janowicz 2012, Elliott 2013, Grossner 2013). ⬈#p7

The addition of "when" to a Linked Ancient World Data landscape that currently focuses on "where" and "who" will not come, however, through the development of a consensus-based gazetteer for globally-recognized period concepts. Even if such a vocabulary could be developed, and even if a majority of data managers were to adopt it for future work (a big "if"), it could not be retrofitted to deal with the diversity of period definitions already present in existing print and digital resources without a significant loss of information. Furthermore, I do not think it is desirable to smooth out scholarly disagreements and differences regarding the definition of periods, for this discussion is a fundamental part of the rich history of scholarship on the ancient world. ⬈#p8

A solution to this problem must therefore take into account the diverse ways in which we use and understand periodization in our scholarship, and it must have the flexibility to accommodate conflicting or changing definitions for periods. At the same time, it should facilitate the discovery of the sorts of information people are likely to want to find using Linked Period Data. Three come immediately to mind: material from the same absolute date-range across datasets employing a variety of periodizations; information from spatially-linked datasets filtered by period; and conflicting definitions of periods proposed by different scholars. In the first case, one might wish to find all records associated with the range 400-300 BC from a series of datasets that use various period terms for the same span (Hellenistic, Classical, Late Classical, Early Republican, Iron Age, etc.). In the second, one might wish to see only records associated with Classical Athens, or only references to the name used for a site in a given period (e.g. Cherson, the name used for Crimean Chersonesos in the Middle Byzantine period), or only records associated with period terms used to describe remains in a particular geographic region. And in the third case, a user might want to compare different temporal spans assigned to the same period term in the same place (Figure 3). So a Linked Data resource for periods should include both the terms used to refer to periods and the coordinates used to describe them in time and space. In order to follow scholarly practice, it should also provide references to the source(s) from which those coordinates were derived.⬈#p9

It would be difficult, if not impossible, to provide this information in a structured form in a top-down gazetteer of period concepts. Ryan Shaw, Eric Kansa and I, however, think that we have found a way to provide it in a bottom-up Linked Data gazetteer of period assertions -- that is, a set of stable references for what authorities say about periods, rather than a thesaurus that seeks to impose consensus about what periods are. This is the goal of the Periods, Organized (PeriodO) project, for which we are actively seeking funding. Beginning with the period assertions collected for sites and regions in GeoDia and a set of structured period vocabularies for a large group of Linked Ancient World Data projects, PeriodO will present period assertions in an JSON-LD schema that includes the term or label used by the source; a date range expressed in Julian Days in scientific notation, with the number of significant digits indicating the degree of precision of the start and end dates; an associated geographic entity (country, region, or site) from a Linked Data gazetteer (Pleiades, Wikidata, GeoNames, etc.); and an authoritative source, also linked to a URI, e.g. from VIAF, wherever possible (Figures 4 and 5).⬈#p10

We plan to house the repository in Github, at least initially, and to mint http DOIs through the EZID system of the California Digital Library. The DOIs will be resolvable to structured, machine-readable definitions for individual period assertions (Heath and Bizer, 2011). Github will also provide the user-management tools to begin to build a community of contributors to the gazetteer: if an authoritative user cannot find a URI for a period definition she prefers, she will be able to add her own. In the first phase of the project, we will create a visualization interface that makes it easier to find periods, see their spatial or temporal extent and the degree of precision of the latter, and compare different temporal definitions for the same period terms or the temporal overlap between different terms. Eventually, we hope to build a reconciliation service and a SPARQL endpoint that will make it easier to implement period assertion URIs and reuse the dataset itself.⬈#p11

As scholars, we do not talk about periods in the same way that we talk about places. With places, we generally tend to agree about the location of those that are well-known, and we generally seek to establish consensus on the location of those that are ambiguous. With periods, although we refer to them in a general way to communicate with each other, we prefer to use precise but idiosyncratic definitions that best fit our own material. We are also careful to cite the sources of those precise definitions, if we did not invent them ourselves: specificity about such matters is a basic tenet of scholarly communication. It seems reasonable, then, to attempt to mirror scholarly practice in the description of periods as Linked Data. The gazetteer PeriodO proposes to create would make room for the disagreement, imprecision, and multivocality that characterizes scholarly discourse, even while providing URIs for structured assertions about periods that include spatiotemporal coordinates and authority citations. This approach will not solve all the problems that face the implementation of period concepts in the Linked Ancient World Data ecosystem, but it does offer a grassroots starting point from which we can begin to explore possibilities. The success of this effort, like the success of Linked Ancient World Data initiatives in general, will depend on the development of a scholarly community of practice, with members willing to implement period assertion URIs in their own datasets and to contribute new period assertions to a gazetteer. Periods are too important to the ways we talk about and conceptualize the past to leave out of the Linked Ancient World Data cloud: it really is about time.⬈#p12