Historians argue that tracing early exchange, trade and uses of plant medicines (materia medica) can elucidate dynamics of drug trajectories in the early modern period. However, information on how these drug trajectories have evolved is hidden in large amounts of heterogeneous historical data. ese data are in dierent formats, languages, genres and stages of digitization, thus creating huge challenges for i) accessing information, ii) presenting aggregations, trends and paerns, and iii) making the research process transparent and traceable for sharing, collaboration and validation. In this paper, we present the historical research platform of the Time Capsule system, from the user's perspective. We illustrate this with an extensive example, to show how the semantic integration of diverse historical data sources into a single linked data knowledge structure in our Time Capsule system has practical applicability for historical research.
Historical research requires analysis of disparate and incomplete data, where the building blocks for creating a narrative are scaered. e researcher aims beyond the most obviously relevant sources, towards discovering hints and unexpected connections in seemingly unrelated documents.
HistoInformatics 2017, Singapore Copyright for the individual papers remains with the authors. Copying permied for private and academic purposes. is volume is published and copyrighted by its editors..
For this reason, we argue that for historians to prot from the full potential of information technology applications in their research, three general challenges should be addressed: access relevant information is scaered across digital sources with limited metadata and content information. e challenge is to enrich semantic metadata and integrate sources in a way that ts historical research methodology. presentation close reading of large and semantic interrelated digital data is not only constrained by the manual eort required, but may also obfuscate larger trends and paerns in such complex data. e challenge is to establish the right form of distant reading that aggregates and summarizes large digital material in a meaningful and relevant way. validation digital information and digital research methods have accentuated the need for transparency. e challenge is to establish methods for tracing and validating the research process and the arguments and claims based on it.
Our work aempts to address these three challenges for digital
historical research in a case study: the trajectories of botanical drug
components in the Low Countries in the early modern period (ca.
16th-18th century). e reconstruction of these drug trajectories
is vital for understanding the global exchange of knowledge and
goods [
In response to this need, Time Capsule1 implements our solutions to information access, presentation and validation challenges. Time Capsule integrates a variety of botanical, linguistic, archaeological and historical data in a single historical knowledge graph. In this way, semantic links re-contextualize initially scaered data, while researchers are provided with a single access point for data 1 hp://www.timecapsule.nu querying and analysis through multiple, interdisciplinary perspectives. Temporal and geographical visualizations present meaningful aggregates for researchers. Finally, system queries and results can be saved and shared in a “time capsule”, which allows users to retrace steps of the search process, to validate the usefulness of results and to share these with other researchers.
A general overview of the considerations, methodology and
implementation of the semantic integration of our initially dispersed
data sources, that form the knowledge backbone of the Time
Capsule system, has been outlined in [
is paper is structured as follows. In Section 2, we present current approaches to metadata and semantic integration of heterogeneous data. In Section 3, we provide an overview of our data sources and methodology for interrelating and enriching disparate data sources. en we discuss Time Capsule’s functionalities, from the user perspective in Section 4. e example, drawn from the history of pharmacy, concerns the reconstruction of a drug trajectory in the 17th and 18th centuries. Finally, in Section 5, we discuss conclusions and plans for future work. 2
Data integration is a common issue in database research, because
of the challenges associated with reconciling disparate database
schemas and semantics, so as to support comprehensive data
querying. An obvious solution is the integration of all data in a single
repository, the so-called in-advance approach or data warehousing
[
For cultural heritage databases, dierent metadata standards
are currently applied, both within and across dierent institutions.
For example, bibliographic data is generally described using the
MARC21 standard [
Aempts for integration of diverse metadata schemas include
mappings across schemas [
Another important challenge lies in data integration for access
beyond warehouses. To address such a challenge requires an
approach that allows integration and use on the Web of disparate
data. Linked Data came as a response to this challenge [
A detailed discussion on using linked data for cultural heritage
is provided by Hyvo¨nen [
In Time Capsule we adopt a linked data approach for data source
integration. According to Hugge [
In our approach, as also illustrated in Figure 1, a domain ontology formally dening our set of concepts (e.g. drug components, plants, reference sources) and respective properties and relations is central. e individual data sources are then converted in RDF and key concepts are mapped to our domain ontology. In the conversion process, we semi-automatically identify ambiguities and spelling inconsistencies, lter out certain non-relevant information (e.g. curator information) and enrich the data with explicit relations among synonyms, term language, explicit temporal information and geographical coordinates information. e laer are semi-automatically acquired by geographical databases, such as
GeoNames2, or historical geographical resources, such as the list of the Dutch East India company trading posts3.
In this section, we rst give an overview of our data sources and their principal contribution to our knowledge structure and then, we discuss our domain ontology and the particular challenges related both to multi-disciplinarity and the historical nature of the domain that we had to address in designing it. 3.1
In Time Capsule, we have semantically interconnected a variety
of pharmaceutical, botanical and linguistic knowledge resources,
primarily in the form of structured or semi-structured data. An
overview of our current data sources is given in Table 1. In our
system, these are also connected to relevant external linked data
sources, such as DBPedia [
e Pharmaceutical-Historical esaurus, developed by the Foundation for Pharmaceutical Heritage, covers terminology related to botanical drug components, such as names and spelling variants of the various botanical remedies, plants of origin and parts of the plant used (e.g. seed, root) that were used in medicine in the past. e information is derived from pharmacopoeias, which are ocial pharmaceutical handbooks that apothecaries on a municipal level were obliged to use (issued between 1636 and 1795). ese manuals dictated to apothecaries which drug components they were supposed to have in stock and the basic medicinal recipes they had to prepare with these. e thesaurus contains the lists of required botanical drug components from 10 pharmacopoeias. It also contains documentation information, such as the curator’s name and various notes about the entries. It was developed using a proprietary thesaurus management tool4 which internally stores the data as an SQLite database.
e Economic Botany Database of Naturalis Biodiversity Center encodes botanical information about approximately eight thousand plants in MS Excel format. e principal objective of the database lies in classifying information about uses (e.g. as dye, remedy, or food) of a given plant species. e data also includes botanical 2hp://www.geonames.org/ 3hps://www.vocsite.nl/ 4MultiTes Pro taxonomic information, geographical origin and common names (English, Dutch and other), including descriptions of the physical specimen at Naturalis.
e data from National Herbarium of the Netherlands of Naturalis Biodiversity Center consists of three herbaria (the Wageningen, Leiden and Utrecht University collections), amounting to 2.8 million records of plant specimens. In addition to botanical plant name, family and genus, the National Herbarium provides information about a specimen’s geographical origin, such as longitude, latitude and elevation. In the Time Capsule linked data structure, the National Herbarium provides the canonical, modern botanical name and taxonomic classication for a given plant species and the basis upon which other plant mentions in other data sources are resolved.
e Snippendaalcatalogus of from the Hortus Botanicus os Amsterdam encodes botanical information about Johannes Snippendaal’s botanical garden in Amsterdam in 1646. is data source provides crucial historical botanical nomenclature and taxonomic information and mappings of Snippendaal’s pre-Linnaean names, but also obsolete scientic names from the 19th century, and current scientic names.
e Boekhouder-Generaal Batavia is a mySQL database developed at Huygens ING, with information referring to the trade of commodities, as found in the accounting books (Boekhouder-Generaal) of the Dutch East India Company.
RADAR [
Finally, the Chronological dictionary [
e Time Capsule domain ontology formally denes the concepts, their properties, and interrelations of the notions salient to our case study of drug trajectories. Nevertheless, it has been designed in such a way, that it should be extensible and applicable to other domains related to the so-called Republic of Materials, or historical study
Artemisia absinthium Artemisia absynthium Absinthium incanum, foliis compositis latiuscule multidis Gemeen alsem in the exchange and appropriation of materials and knowledge about them, in the early modern period. Currently, the principal concepts in the ontology revolve around three notions: Naturalia, or substances derived from nature, such as animals or plants and their parts; Drug Components and their properties; and Reference Sources, or sources of information for data, properties and relations stated in our instances, that originate from text documents or physical specimen references.
Apart from designing the ontology to be extensible and re-usable as much as possible based on input from experts on history, biology and botany, we had to address three particular challenges related to our domain: (i) variation, (ii) ambiguity and, (iii) scientic evolution of terms. Variation is a common phenomenon in scientic terminology (even more so in historical text sources), where spelling conventions are still uncommon and the scientic nomenclature is not yet standardised. Figure 2 illustrates an example of remedies found in our sources that may, or may not, refer to the same drug component. Ambiguity is related to variation and is exacerbated by underspecication, vagueness and uncertainty in historical sources. For example, a reference to Radix Chelidonii or Chelidonii does not clarify whether the remedy originates from the plant called Chelidonii majoris or Chelidonii minoris in the past. It might also be the case that a drug component is associated, under a single name, to dierent plants or plant parts (e.g. root or bark), in dierent reference sources. Finally, scientic evolution of concepts is a particular challenge, that is mostly related to the domain of the history of science. e evolution of human knowledge and science from e.g. the early modern period to modern times, inherently aects both the concepts and the terminology that is used, as well as the organisation and classication of knowledge itself. In designing our knowledge structure we therefore had to account for multiple “versions” of knowledge about the world at dierent moments in time.
Figure 3 illustrates our approach to the challenge of scientic evolution. e concepts associated with Naturalia (in dark red), either Plantae, Animalia or Mineralia, originate from the classication of knowledge in the early modern period, whereas the taxonomy Kingdom > Order > . . . Plant Species is based on the modern botanical classication (in light green). In our ontology, we model both knowledge versions in the same knowledge structure and aempt to provide mappings/links, where possible, across time, so that a historian or other user can access all relevant data for longer time frames.
In order to address variation and ambiguity, we have adopted an ontology structure, as illustrated in Figure 4. is structure (i) separates statements about a given concept in a reference source from the concept instance itself; (ii) does not resolve ambiguity, but rather follows a fuzzy classication, which allows an instance of e.g. a drug component to be associated to multiple plants of origin. In this way, we aempt to stay faithful to any information stated in a reference source, while reducing as much as possible the inherent bias in the knowledge modeling of our system and leaving ambiguity and variation open to a researcher’s interpretation.
In the example in Figure 4, the instance of a plant concept (Dictamnus albus L.), relates to the mention Dictamnus albus L. in the Snippendaal Catalogue and the National Herbarium references, via the property hasVariant. According to the Snippendaal mention, this plant species, Dictamnus albus L. isReportedToProduce two remedies, one drug component originating from the root of the plant, Radix Fraxinella, and one originating from the bark, namely Cortex Fraxini. In this example, we can also see that there is a physical specimen of the plant mentioned in the National Herbarium, a specimen originating from the University of Wageningen collection, with ID WAG 1255711 (and respective geographical coordinates of the specimen origin, which are not included in this gure).
e domain of application for this work relates to research in the history of pharmacy. In particular, we focus on cultural heritage data sources that reveal (parts of) trajectories of botanical drug components in the Low Countries, from the sixteenth century, when natural drug components from the East and West Indies started penetrating Europe, until the introduction of chemical and synthetic drugs, roughly in the mid-19th century.
Within historical research, drug trajectories denote
developmental processes of remedies, such as growing economic importance of
a drug because of a steady supply; scientic interest shis because
of ongoing research in therapeutic ecacy and side eects; drug
acceptance or rejection by (segments of) society for nancial,
religious, ethical or other reasons. In other words, drug trajectories
are specic aspects of the history of drugs, which can be studied
diachronically, i.e. tracing one such aspect over a longer period of
time) or synchronically, i.e. studying/comparing the interplay of
different spatial contexts at a given point in time. Studying the history
of pharmacy through the perspective of trajectories of individual
drugs is a fairly recent approach that allows researchers to trace
larger developments in the history of therapeutics. For example, the
adoption of a given drug component may only partly result from
its therapeutic qualities. Equally important in this process may
be non-medical factors, such as public acceptance, marketing and
availability, factors revealing the dynamics of the medical market
during the historical time frame under consideration [
In our work, we are particularly interested in the trajectories
of drug components from the East and West Indies and how these
relate to networks of trade and knowledge circulation in the early
modern Low Countries. In this period, natural history is considered
the central focus of intellectual and commercial activity [
Time Capsule is meant to provide the aggregated data required for new information to surface which is dicult for an individual researcher to produce by means of conventional historical research alone. Moreover, our system is aimed at assisting the researcher in analyzing these vast amounts of data, by providing spatio-temporal visualizations that give clues for a meaningful interpretation. In this way, existing historical narratives will be enriched.
In designing an interface and an online platform for information
access, one should consider that the potential users may not be
domain or system experts. It is oen dicult for non-expert users
to perform queries, either because they are unfamiliar with the
required terminology, or with the available information and the
underlying data model. Our solution to this issue lies in providing
two querying strategies in our Time Capsule research platform
interface: one that supports faceted exploratory search and browsing
of information by means of links, and keyword auto-completion
suggestions; and one that supports the creation of ad hoc queries.
e exploratory search mode is intended to engage a wider audience
and reveal to both expert and non-expert users the content and
structure of the underlying data. e user interface allows easy
construction of ad hoc queries, by means of a user-friendly query
wizard, which supports the creation of RDF SPARQL queries [
To illustrate the usefulness and user-friendliness of our system, we discuss an extensive example, where a researcher focuses on information for a drug component called Asafoetida, a resinous gum that was used in pharmacy from the 16th century onwards. Due to its strong smell, it was known as Devil’s dung in many European languages. In this example, we want to know how its historical trajectories evolved over time.
e rst and most obvious way of tracing information about Asafoetida in the Time Capsule system is by searching for it in the Drug Components tab (see Figure 5). e Drug Components tab is part of the faceted search interface which the Time Capsule platform provides for browsing, in this case, all system knowledge related to drug components. With the auto-complete function in the search text box (on the le hand side of the screen, indicated by the magnifying glass), several spelling variants for this drug component pop up (the list with red bullets below the search text box). Clicking on any of them will show the information found in the reference sources for this drug component: six dierent spelling variants; four Reference sources that mention this substance; a map that highlights the Publication location of the References; and a list with a total of 25 references to this drug component, which are all derived from one data set, the Pharmaceutical-Historical esaurus.
At this stage, no information is shown in the Produced By Plant header, which means that there is not any explicit information, i.e. relation, in our knowledge graph about the plant species from which the drug component Asafoetida is derived. However, the system provides other options to proceed. As illustrated in Figure 6, the query for Asafoetida also gives results in the Naturalia tab, namely the part of the faceted search in the Time Capsule platform that allows browsing and search of information related to all plant species information in our system. We discover in the Naturalia tab section that Asafoetida is also a spelling variant for the plant from which the drug component with the same name derives: Ferula assa-foetida. e result for this query shows several spelling variants for the plant, which are similar but not identical to the spelling variants for the drug component Asafoetida. Unsurprisingly, there are not any drug components for this plant mentioned under the Produces Drug Components header: we already saw that the explicit relation/link between this plant and its respective drug component is absent.
However, we discover additional information next to the Name Variants, namely a list of reported Uses. ese data, originating from the Economic Botany Database, provide a number of possible contexts in which we might encounter Asafoetida. Apart from its use in medicine, Asafoetida is used in food, in spiritual cures and as an essential oil. Moreover, above the Name Variants we can nd a text description about the plant genus and a photo, both resulting from linking our Time Capsule data to DBPedia linked open data. DBPedia informs us that the Asafoetida plant genus originates from Central Asia. is makes the historical use of Asafoetida in European pharmacy all the more interesting, because it would have required an intercontinental supply chain.
e information from DBPedia is corroborated by the integrated data in Time Capsule, as visualized in the map on the same page with results, illustrated in Figure 7. Apart from some expected locations in Europe, and one undened pointer in the Dutch Caribbean, we also see one location pointer in India, where we would expect a Natural Distribution location, i.e. location of plant origin mentioned in our botanical specimen data.
At this stage, we need to explore in more detail the Asian connection that we came across in several of the data collections in the Time Capsule system, and investigate whether we can discover any additional information about the complex commercial trajectory of Asafoetida in the past. To do so, we can use once more the query Asafoetida to browse in the Cargo & Trade tab. is is another part of the faceted search in the Time Capsule platform that allows browsing and search of information related to cargo carried by the ships of the Dutch East India Company in the 18th century. In this case, we want to investigate whether Asafoetida is found in any of these cargo records, as a trade item.
e results are illustrated in Figure 8. e green leaf next to the cargo icon in the search bar indicates that we are indeed dealing with a vegetable substance, namely a cargo item classied as a natural, plant substance. is leaf symbol is the result of the integrating and enrichment process that the respective Boekhouder-Generaal Batavia data set underwent in Time Capsule, that specically marks cargo items such as Naturalia (among other classied materials recognised in this process). e query result shows 61 journeys where Asafoetida was part of the cargo. We know this result to be somewhat inaccurate, because journeys are sometimes listed more than once if more than one departure/arrival location is mentioned in the original data (e.g. a journey departing from Kharg in Persia is listed twice, once departing from Kharg and once from Persia).
ese results show that most ships carrying Asafoetida follow one of two routes: either within the South-East Asian region, where only the place of destination is a Dutch selement, or between South-East Asia and the Dutch Republic. In other words, the Trade & Cargo results provide convincing data for further analysis of Asafoetida’s main supply route. e substance was apparently purchased in a region not controlled by the Dutch (India), then shipped to a Dutch emporium (usually Sri Lanka, or Batavia), and then shipped to the Dutch Republic. Details of one of these 61 journeys, as illustrated in Figure 9, departing from the Coromandel coast in India to Jana on Sri Lanka on 31 August 1767, reveals that this is an example of the South-East Asian trade route described above. e ship carried, among other things, a modest amount of 167 pounds of Asafoetida: modest, because other ships carried up to thousands of pounds of this substance.
We now know that Asafoetida was shipped regularly from SouthEast Asia to the Dutch Republic by the Dutch East India Company, in substantial quantities, over the course of the 18th century. We can now try to determine to what extent there was demand for this substance, not by browsing, but by using the Query Machine, (see second tab in Figure 5). e Query Machine implements, on the
Time Capsule interface, the part where the user may form ad hoc RDF SPARQL queries, using a user-friendly query wizard.
In the Query Machine we formulate the custom query: “List All Reference Source(s) that mention Asafoetida”. e system gives six results, all of them from pharmacopoeia references, among which two editions of the Pharmacopoeia of Amsterdam (1636 and 1660). We know this result to be incomplete, because we know from our domain expertise that Asafoetida is mentioned in every early modern pharmacopoeia. For this reason, we would expect at least 10 pharmacopoeias to surface in our results, since the Pharmaceutical Historical esaurus data set contains data from these pharmacopoeias. However, although the results are not as precise as we thought, they still provide us with enough input to pursue our research further: the results indicate that apothecaries were supposed to use Asafoetida in Amsterdam in the mid-17th century. Since we know that the pharmacopoeia of Amsterdam was the ocial guide in many cities, our expectation is supported that both supply and demand were substantial.
In this example we jumped from one tab to another, using the results acquired at each step as input to extend our search strategy and knowledge about our main subject. In the process, we stumbled on deciencies that inevitably show up in the data, but we never reached a dead end. Of course, the heterogeneous breadcrumbs we pieced together do not constitute a complete narrative about the history of Asafoetida. ere are many more dimensions than can currently be traced in our system, such as the various Uses we came across, other than as a medicinal substance. But the fact that these bits and pieces could be found within one system has helped us. It allowed us to rapidly assemble very dierent kinds of information in a way that sustains the intuitive strategy of the researcher. It also showed connections that were not in the original data sources, and hence would not have surfaced if a researcher had sied through the individual data sources manually. In that way, the Time Capsule system provides a substantial improvement of the researcher’s digital toolkit, allowing him to save time while retaining an exploratory search strategy, and to stay in command of the research process, while outsourcing time-consuming parts to a digital search mechanism. e reduced eort in piecing this information together is made possible by the substantial eort of high-quality data integration, but this needs to happen only once to support many dierent explorations. 5
In this paper, we presented the Time Capsule system and research platform which contributes to digital history in several ways. It oers a new historical data model and knowledge resource and a novel approach for historical researchers. e system is built in such a way that it may not only be useful for researchers who are interested in early modern drug trajectories from dierent angles, but also for researchers interested in building, using and sharing similar research infrastructures.
rough semantic data integration, historians can query and analyze early modern historical data as a single knowledge graph and contextualize relevant information from various sources and aspects (botanical, linguistic, historical, ethnological, archaeological). Another important aspect in contextualisation lies in making explicit the relation of data to the dimensions of space and time, thus allowing the virtual spatio-temporal “reconstruction” of knowledgetransfer trajectories. Combining knowledge contained in dierent data resources requires a substantial research eort, due to the diversity and ambiguity of the original data. If dierent data sources are available to the researcher, but they are not interlinked, then the added value of having access to more data is rather limited, because the researcher has to continuously perform mapping and disambiguation when switching between resources. Domain knowledge modeling in the Time Capsule ontology and data source integration in an interlinked knowledge graph produces richer query results, that are less time-consuming to obtain; with transparent and retraceable data provenance; more trustworthy for a user than their original data constituents due to context; and thus beer suited as building blocks for constructing a historical narrative.
Regarding the analysis of trends and paerns across data sets, distant reading methods stand or fall with the completeness and quality of the underlying data. Global paerns are hard to trace, given that the available data sets are limited to specic organizations that gathered information relevant and available to them. Obtaining and meaningfully adding information on the quality of digitization and data integration is far from trivial, as certain quality issues may remain hidden under the surface and there is currently lile consensus on a proper methodology of digital sources and digital tool criticism. at is why domain expertise is essential in both digitization and integration stages. However, for data constituents that provide a fairly accurate and complete image (e.g. when an organization like the VOC applied a uniform, sustained method of describing the plants destined for shipment), the links among the data allow analysis of paerns beyond what would be possible for an individual researcher to query by himself. As more data sets are added, in combination with information about their completeness and their digitization and integration quality, the potential for paern and trend analysis increases signicantly.
In the current version of our system, structured database sources have been processed and integrated, and made accessible via a graphical interface for querying the aggregated data. We are nalizing a user study regarding the interface, and a detailed report on the data modeling and integration process which will be published separately. Moreover, additional work is required in optimizing the various data visualizations, which are key to understanding such complex data structures.
Future work includes the extension of our system to additional collections. e Time Capsule knowledge base can be easily enriched with mineral and animal drug components and other substances beyond the domain of Naturalia, so as to cover extensive information about the Republic of Materials. e platform oers many possibilities to connect to other data sets in the future, such as the Sound Toll Registers, that could provide us information on how, for example, Asafoetida found its way across Europe aer it came from Asia to the Netherlands. Finally, we plan to investigate ways to connect further with unstructured data sets. Historical documents provide a wealth of additional information, as well as valuable contexts for interpreting structured data. We plan to use our existing knowledge resource for processing and linking unstructured historical text to our data model.
is work was supported by the Netherlands Institute for Scientic Research (Project Nr. 314-99-111).