The paper discusses the inherent potential of the Semantic Web and its related technologies for humanities research. The focal point lies on the extraction of semantic relations from heterogeneous XML based scholarly corpora using a webservice based infrastructure (XTriples). Especially the creation of methodologically distinct semantic corpora stemming from data sets originating in the humanities will be discussed. During this discussion, questions of modelling, linking, and visualising data from the humanities will be tackled as well. Finally, opportunities for further analysis and visualisation of semantically modelled data in the humanities are exemplarily presented and discussed.
<correspDesc key="686" cs:source="#SOE20"> <correspAction type="sent"> <persName ref="http://d-nb.info/gnd/118540238">
</persName> <placeName ref="http://www.geonames.org/2812482">;
Weimar </placeName> <date when="1793-12-05">
5.12.1793 </date> </correspAction> [...] </correspDesc> (Inherent semantics in XML: Goethe sends a letter from Weimar 1793. RDF subjects are implicit in the cs:source and ref attributes. Predicates are implicit in the date tag and the type attribute. Objects values are implicit in the correspDesc tag, the placeName and persName text nodes and the when attribute.)
Most approaches in the humanities base their e orts in this respect on the
Resource Description Framework (RDF). RDF enables scholars to formulate
and annotate semantic statements in a clear and concise way by utilising the
triple notation: Subject|Predicate|Object. RDF's particular strong point lies
in interlinking, merging and analysing (reasoning) essentially distinct data sets.
RDF in general and webservices like XTriples in particular can also be used as
a means to bridge the gap between so called \altruistic and egoistic modelers"
([
Weimar is_a sends dates_to sent_from is_a has_longitute has_latitude
1793-12-05 ;
(The same example as above in simpli ed turtle notation: First column shows subjects, second column shows predicates, third column shows objects)
Although the formal development of structures and metadata within texts and other objects of research in the humanities can be considered rather advanced, the development of semantics and semantic annotations is still lagging behind. Admittedly, toponyms, personal names or work titles are often annotated nowadays. These annotations, however, con ne themselves mostly to pointing out the occurences of a speci c entity within a data set.
The annotations used within data repositories in the humanities often fall short of the potential o ered by Semantic Web technology in general, and more speci cally, by concepts like linked open data (LOD). LOD o ers the opportunity to join isolated data sets. Thereby, researchers hope to create new, interdisciplinary, and stimulating viewpoints towards established topics.
A key factor in this regard is that LOD and RDF extend established
standards from the Digital Humanities, e.g. TEI-XML, with additional uses,
terminologies, and metadata schemata (see [
Currently there exists a wide gap: On one side of the divide lie the humanities' numerous repositories containing data with great semantic potential, on the other side lie the Semantic Web technologies and data models, which could open up new methods of analysis. Although some concepts, methods, and also tools exist for transforming TEI-XML into RDF and vice versa, these are for the most part overly complex, partially obsolete on a technical level, only implemented as prototypes, or are highly specialised for a speci c type of transformation scenario.1
Whereas the computer sciences regard the technologies and concepts
underlying the Semantic Web as more or less fully developed and applicable (see
[
To facilitate this kind of semantic extraction, the Digital Humanities departement of the Academy of Sciences and Literature Mainz (www.digitale-a kademie.de) has developed a generic webservice called XTriples (http://xtri ples.spatialhumanities.de).
XTriples makes it possible to extract RDF statements out of any HTTP based XML repository using a simple con guration based on statement patterns. The webservice's guiding principles are: { Generic ) works on any XML { Simple ) easy to con gure { Powerful ) for building complex statements with proper ontology support { Flexible ) returns several formats { RESTful ) uses http for request and response { Location-independent ) freely accessible via the WWW { Platform-independent ) server-side processing { Customisable ) adaptable to a project's needs
The webservice is able to crawl any XML le, extract semantic statements and subsequently return them in a speci ed format (cf. g. 1 on p. 58).
The webservice can be used with direct POST, form-style POST or GET requests. The required statement patterns are passed to the webservice in the form of a simple XML con guration based on XPATH/XQuery expressions. The triples are constructed based on the instructions given in the con guration (cf. listing 3, the documentation under http://xtriples.spatialhumanities.de/ documentation.html as well as the exemplary data sets under http://xtripl es.spatialhumanities.de/examples.html).
During the extraction it is possible to reach beyond the boundaries of a speci c XML repository and include external XML resources or data units from these resources on-the- y. This way norm data from the German National library' s authority le (GND), data from DBpedia, or any other third party can be included. Furthermore, the con guration can be used to ne tune the quantity, the type, and the granularity of statements the webservice is supposed to extract. Listing 3: Sample structure of an XTriples con guration <xtriples> <configuration> <vocabularies> <vocabulary prefix="tei" uri="http://www.tei-c.org/ns/ 1.0"/> <vocabulary prefix="cs" uri="http://www.bbaw.de/
telota/correspSearch"/> [...] <vocabulary prefix="foaf" uri="http://xmlns.com/foaf/
0.1/"/> [...] </vocabularies> <triples> <statement> <subject>//tei:correspAction[@type="sent"]/
tei:persName/@ref</subject> <predicate prefix="rdf">type</predicate> <object prefix="foaf" type="uri">Person</object> </statement>
[...] <statement> <subject>//tei:correspAction[@type="sent"]/
tei:persName/@ref</subject> <predicate prefix="rdfs">label</predicate> <object type="literal" lang="de">//tei:correspAction
[@type="sent"]/tei:persName/text()</object> </statement> </triples> </configuration> <collection uri="http://correspSearch.bbaw.de/api/v1/ tei-xml.xql?correspondent=http://d-nb.info/gnd/ 118540238&startdate=1793-01-01&enddate=1808-02-02"> <resource uri="{//tei:correspDesc}"/> </collection> </xtriples> (Below <collection> the XML data that should be processed by the service is congured. Below <vocabularies> it is possible to con gure the RDF vocabularies that should be used. Below <triples> the <statement> patterns that contain the subjects, predicates and objects are con gured. The notation will output the following in turtle syntax: cs:personID rdf:type foaf:Person ; rdfs:label "Johann Wolfgang von Goethe"@en .)
The implementation as a webservice has the clear advantage that the user does not need to install any software besides a web browser in order to translate data from XML to RDF. XTriples can also be used as an (external) RDF interface (like a proxy) for one or more XML based repositories. The prerequisite for this kind of usage is simply a repository's availability via HTTP.
The webservice can return the following formats: RDF/XML, Turtle, NTriples, NQuards, Trix, JSON, SVG, and custom XML for debugging purposes.
The result of an XTriples extraction is thus available in a wide variety of RDF-serialisations (cf. g. 2 on p. 59). As seen above, besides the purely RDF based format, a repository's semantic relations can also be written to an SVG le or piped to other Semantic Web tools.2
XTriples was developed in the context of the long term research project Deutsche Inschriften (German Inscriptions, currently carried out by six of the eight German Academies of Sciences) together with the project Inschriften im Bezugssystem des Raumes (Inscriptions in their Spatial Context, funded between 2012{2015 by the German Ministry of Education and Research). The software is released on Github (https://github.com/spatialhumanities/xtriples) in a stable version under MIT license and comes with a detailed technical documentation (http://xtriples.spatialhumanities.de/documentation.html). The XTriples webservice is currently used in di erent projects associated with or located at the Academy of Sciences and Literature in Mainz. Section 3.1 will focus on the the extraction of RDF statements from XML in a use case from the eld of spatial humanities. Section 3.2 will exemplify the webservice's capabilities within the contexts of three further digital humanities projects and thus illustrate the webservice's exibility in di erent project environments and its adaptiveness to their varying research questions. 2 It is for example possible to pass the data to the RDF-to-SVG webservice (http://www.rhizomik.net/html/redefer/rdf2svg-form) or to the RDF visualisation library d3sparql (http://biohackathon.org/d3sparql). The XTriples URL can also be passed as parameter to Visual RDF, e.g https://graves.cl/visualRDF/?url=xtriples.spatialhumanities.de/extra ct.xql?configuration=YourXTripleConfig.xml.
The Projects \German Inscriptions Online" and \Inscriptions
in their Spatial Context"
The long term research project Deutsche Inschriften (DI) is a joint undertaking
of six German Academies of Sciences and the Austrian Academy of Sciences.
The research focuses on collecting, editing, and interpreting medieval and early
modern Latin and German inscriptions. They often occur in conjunction with
gurative elements or spatial as well as architectural features. The inscriptions
themselves are mostly in medieval Latin or in historical or regional varieties of
the German language. The geographical area of research consists of Germany,
Austria, and South Tyrol. The inscription records range from approximately 500
AD to 1650 AD [
The Federal Ministry of Education and Research (BMBF) funded project
Inscriptions in their Spatial Context (IBR, 2012{2015) had the aim to combine
and analyse spatial data from terrestrial laser scanning with the epigraphical
data made available by DIO. To achieve this, relevant parts of the epigraphical
data (marked up in TEI-EpiDoc) had to be made available as RDF. This was
done with the XTriples webservice. To achieve this, IBR rst compiled its own
semantic predicates in a project speci c ontology (see [
The IBR project then harvested relevant TEI-EpiDoc records from DIO's
epigraphical database via its REST interface according to guidelines formerly
worked out by the project's participants. Afterwards the harvested EpiDoc
corpus was processed by the XTriples webservice and translated to RDF. The
transformation was very e cient regarding time and labour as well as highly suited
to the project's needs. It was possible to incorporate the project's custom-made
ontology. This high level of adaptability was due to XTriples inherent
customisability via its exible con guration.3 Subsequently the RDF data was loaded
into the project's main working environment, the Generic Viewer [
Using this synthesized data approach, IBR was able to allocate spatial
segments of a church's interior to distinct socio-political groups. Besides this, spatial
areas within the church could be identi ed and annotated with their respecitive
liturgical, ritual, or social function, thereby highlighting \potentially relevant
factors like the social division of the congregation room, procession routes, and
other places of liturgical practices" [
IBR's researchers were able to estimate the former installation or in situ
location of epigraphical artefacts|for example an epitaph and retable endowed
by the canon Petrus Lutern|by means of a Viewshed -analysis [
This use case highlights the applicability of semantic annotations in regard
to historical research within a highly technical environment. But the concept
could also be very useful in regard to more public-oriented spatial visualisations
[
Other Projects Besides the projects DI and IBR, XTriples is currently used by several other academic projects, namely the Regesta Imperii 4 and the Schule von Salamanca.5 The Salomon Ludwig Steinheim Institut fur deutsch-judische Geschichte 6 uses XTriples for a CIDOC-CRM based modelling of its EpiDoc-data. Together with 3 At this stage it would also have been possible to process the data further, e.g. by visualising the semantics with SVG or by further enriching the RDF le with incorporated external sets. 4 http://www.regesta-imperii.de 5 http://www.salamanca.school/ 6 http://www.steinheim-institut.de the Digital Humanities departement of the Berlin-Brandenburg Academy of Sciences (BBAW) a joint e ort is under way to develop a common API between XTriples and correspSearch, the BBAW's decentralized aggregation-tool for metadata of digital scholarly editions of correspondence.
The following list gives a rst insight into some of the diverse use cases of XTriples: 1. Extraction of semantic relations and subsequent visualisation of the familial ties present within the data made available through the Steinheim Institute' s EpiDat corpus.7 The data represents the Jewish cemetery in HamburgAltona.8 2. Extraction of semantic relations from Johann Wolfgang von Goethe's correspondence and subsequent SVG visualisation of a speci c sub-network. The underlying data was retrieved from CMI records aggregated by correspSearch. Norm data provided by Geonames and the DNB's authority le was incorporated on-the- y via their respective RDF interfaces.9 3. Exemplary extraction of semantic relations from correspSearch' s CMI/TEI database and subsequent visualisation of European communication networks implicitly present within the records ( g. 3 on p. 62).10
More use cases demonstrating particular functionalities of XTriples can be found on the XTriples website. The following presentation gives a quick overview of the service's capabilities and modes of operation.11 4
Conclusion
As shown above Semantic Web tools that wish to step up to the challenges posed
by humanities research have to address many di erent topics. When working with
data sets from the humanities the huge diversity found in this eld of research
always has to be taken into account. Generally speaking it is hard to nd even
two digital scholarly editions that are structurally comparable or pose the same
research questions, even though they may be working on the same topic. Any
semantic web tool that should be usable for humanities research and data must
therefore be able to deal with this surprisingly high degree of diversity. It should
not be con ned to a xed set of ontologies, schemata, or output formats, but
permit the highest possible degree of exibility and thereby further \a spirit of
openness shared by all parties involved" ([
Thinking in terms of monolithic software is not enough when it comes to
bridging the gap between modelers with di erent approaches (see [
To summarise, it is appropriate to say that in the same way as the Semantic Web distinguishes itself by simultaneously being distributed and heterogenous and, at the same time, connected and analysable, the tools and the researchers trying to utilise Semantic Web technologies for the humanities have to be highly exible in their approaches, too. It may be worthwile to change the focus from an all-in-one to a more project and microservices oriented approach that scales much better to the semantic potentials concealed in the many di erent data repositories from the humanities that are available online.