The German Biography Portal "Deutsche Biographie" is a joint effort of the Historical Commission at the Bavarian Academy of Sciences and Humanities and the Bavarian State Library and supported by cultural heritage institutions to develop a historical and biographical information system for the German-speaking world. It includes digital full texts of more than 48.000 articles about persons and families of two biographical dictionaries and indices from associated institutions. We will describe our objectives: textencoding, identifying individuals and places in authority files and aggregating further individual/biographical information from freely available, persistent, scientific and source-based websites and databases. The portal offers an entry point for historical biographical research. The potential of it lies in coordinated biographical data management and integration. The common database is gradually enlarged in a collaborative and modular manner together with partners in Germany and Europe. We will discuss on how the data base of scattered information can be managed.
Biographical dictionaries constitute a unique source of historical knowledge. The articles not only inform about its subject, but they also give insight into the authors, their questions, methods, intentions, biases, and the zeitgeist. The factual knowledge expressed can be regarded as metadata. And, as we will show in this article, it can be successively derived from text, expressed using semantic web-inspired ontologies and hopefully reasoned upon. We achieved the best results by interlinking recognized entities (names, places). And we will try to pursue this approach further (works and similar subjects). Our efforts relate to similar prosopographical projects. The biographical portal of the Netherlands successfully merged and enhanced several digitised historical biographical dictionaries as well as current online resources. 1 An API 2 allows automated request of database entries. A currently still ongoing project is described online and focuses on metadata and relations between individuals. The portal does not produce biographies on its own. The Oxford University Press (OUP) set up an index combining individuals from the Oxford Dictionary of Biography with the National Portrait Gallery and National Archives content (images), as well as indexes of several monographs and dictionaries they published. The OUP is constantly extending their fee-based online database by 1 cf. http://www.biografischportaal.nl/about and the interchange format description. http://www.biografischportaal.nl/about/biodes. 2 http://www.biografischportaal.nl/about/bioport-apidocumentation.
about 200 biographies each year and published two supplement volumes, covering people who died 2001-08. 3 Available online is a search gateway "Oxford Index" based on identified names linking to 16 selected resources. 4 There are no known projects outside OUP which reuse these identifiers apart from Wikipedia/Wikidata. 5 The Australian Dictionary of Biography 6 digitised their historical material as well. Their efforts concentrate on both the digital continuation of the dictionary and the enhancement of digitised articles. 7 The dictionary reconceived their publishing plan: new articles are still written, gaps are identified and subsequently filled. The articles now are provided with individuals mentioned, topics and nationality by birth using controlled vocabularies. Articles from a few other national relevant sources 8 are also integrated, obviously by manual redaction. There are only internal identifiers, no metadata is published separately. Users can aggregate several statistics on the database online.
The Historical Dictionary of Switzerland (HLS) started in 1989 and published both online and in print from the start (cf. Jorio 1990 and Jorio 2000). The articles link to the authority file Gemeinsame Normdatei (GND) and the Virtual International Authority File (VIAF) where appropriate, missing identifiers are added to the GND resource. 9 Finally the tripartite complex Wikipedia, DBpedia and Wikidata must be mentioned. The DBpedia arose out the Wikipedia-dictionary as a means to conceptualize facts and assertions in a structured way. In addition the practical requirements to manage obvious facts scattered in different language versions gave birth to the Wikidata effort. Wikidata manages merely simple assertions on articles and struggle with the problem that facts and ontological structures are not always translatable. 10 The digitisation of the NDB/ADB fits into the field. The latest volumes (V-Z) are not finished, all volumes in print are digitised and published online after an embargo period of 2 ½ years. The demand for data in science is met by creating and publishing structured metadata and offering link protocols.
The precursor of the "Neue Deutsche Biographie" (NDB) originates in the 19th century. The "Allgemeine Deutsche Biographie" (ADB) was compiled and printed in 55 vol., between 1875 and 1912. This dictionary was embedded into a programme to shape the national identity on the basis of a German culture and language. The concept was to reinforce a kind of federated cultural nation on the basis of protestant liberal middle-class milieus including opponent and minority positions (Hockerts 2008, 238).
The NDB was outlined towards the end of World War II in order to renew the former dictionary ADB and in the preface dedicated to the German people as a whole ("das dem ganzen deutschen Volke gehören soll" Vorwort, NDB I, 1953, p. VI). It was conceived to cover distinguished individuals of German language and culture and was not confined by national borders. The culturally grounded concept included members of the German Volk (in Austria and shortly before detached territories) and foreigners with cultural influence in Germany (Vorwort, NDB I, 1953, p.VII f., in detail Hockerts 2008, 244-254).
As of 2015 25 volumes in alphabetical order up to "Tecklenborg" have been published. Three volumes ("Tecklenburg -Zyrl") are planed to be released until 2020.
The NDB gives concise, thoroughly prepared biographies of deceased persons who have had a significant impact on developments in scholarship, literature, arts, politics, economics, social life, and technology (Hockerts 2008).
A GND-Beacon is provided: http://www.hls-dhsdss.ch/pnd/BEACON-PND-HLS.txt.
The NDB is regarded as an authoritative biographical dictionary for all regions in which German is spoken ("deutscher Sprachraum") and German culture is prevalent (Hockerts 2008, Kraus 2010).
The NDB covers the period from the early middle ages down to the present and is arranged alphabetically. The 25 published volumes, contain about 22.000 articles, roughly 19.000 are biographies on individual persons and 3.000 cover families. The articles include detailed information on genealogy (cf. Ebneth 2012), selective lists of works as well as secondary literature, and references to portraits. In total about 8.000 different authors, often distinguished experts in their field, have contributed articles.
The NDB is edited by the Historical Commission at the Bavarian Academy of Sciences and Humanities, Munich, under the guidance of the editors in chief, who are members of this commission. 11 The publishing house Duncker & Humblot, Berlin, printed and distributed all volumes, each containing about 800 articles on 830 pages and an index of names. The NDB is mostly used in public and scientific libraries in Europe and Overseas. Each volume is prepared by an editorial staff of five historians each of whom is responsible for a particular subject area. The staff selects the people to be included in the NDB, appoints qualified authors, and edits the articles for printing. The preparation process builds upon an internal database which is continuously extended by systematic examination of online resources, monographs, periodicals, newspapers, obituaries, bibliographies, editions and exhibitions. Each volume covers a selection of 8.000 names in the internal database, the editor responsible for a specific field (humanities, sciences, literature, arts, politics, business, medicine) creates a preselection which is discussed with selected relevant experts. The author is chosen by the editor and the article undergoes an informal peer review process in the editorial office and a revision in order to fit the criteria of style, content and completeness.
A strict structure is imposed on all articles (cf. Redaktion der NDB 2009):
1. Full name, occupation, date and place of birth, date and place of death, tomb, religious denomination 2. Family (genealogy) 3. Career, achievements, critical evaluation 4. List of selected works 5. List of sources and secondary literature 6. References to portraits (but no illustrations) 7. Name of the author.
The "making of" www.deutsche-biographie.de rests on three pillars: (1) the digitisation of ADB and NDB with the ongoing edition of biographies, (2) a joint index of ADB and NDB and (3) the interlinking and integration of distributed sources.
A dictionary of notable individuals and families is generally understood as a principal source of biographical information and should provide a low threshold of accessibility. At first the pages of the older series ADB and subsequently the NDB had been scanned and put online. The index of names was already compiled for the ADB in an additional volume in print. It was and is enlarged by all the names from the indices of each new volume of NDB. The index of names, professions, dates and pages of occurrence accompanied the digital images.
A relational databases back-end served as a comfortable means to find the specific pages.
In a second step the full text was digitised, encoded in XML (partially TEI12 ) and the structure reintroduced. In parallel each name was aligned to the bibliographical authority file Personennamendatei (PND, in 2012 extended to GND, cf. Pfeifer 2015). By exploiting structural tagging and indexed information the cross linking of articles and index entries could be realized.
The third stage introduced faceted filters and map based search abilities.
A constant interest of the editorial board is to extend the knowledge base, to search for reliable biographical information on historical personalities and families. In 2009 a decade after initial thoughts and plans an international cooperation started (Menges/ Ebneth 1998, Ebneth 2009, Ebneth 2010). The Biographical Portal brought together the index entries of several national biographical dictionaries and made them searchable in a unified way (cf. Gruber/Feigl 2011).
In addition the NDB gained substantial support through cooperation. Relevant cultural heritage institutions delivered identifiers and the meta data could be drawn from the authority file GND.
A basic form of cooperation with over 100 websites is realised by cross-linking on the basis of authority file identifiers and a suitable interface (BEACON). 13 By publishing a list of GND identifiers and a concordance to subpages on a website they become interlinkable through aggregators. 14 These concordances could as well be created by a third party researcher.
The first database online already offered selections along groups of profession.
To ease access to all relevant scientific sources it is required to improve and systematize linking. The NDB besides others adopted and promoted the use of authority files and identifiers in historiographical research projects (cf. Akademienunion 2009). Following prominent proponents like BSB and Wikipedia the NDB promoted simple protocols (BEACON) to offer and share lists of identifiers and concordances. 18 With BEACON everyone can aggregate lists of links on an individual basis automatically.19
In late 2011 the Historical Commission applied succesful for consulting in an EU-funded project Linked Open Data (LOD2, Riechert 2011). 20 Together with experts from the Leipzig AKSW21 the metadata were translated into a semantic web ontology expressed in RDF-XML (Brümmer 2011). This prototype22 was based upon the ontology-schemata developed for the authority file GND and comprised roughly 2.7 million statements.
15
The GND is curated by the German National Library and widely used by german, austrian, and swiss libraries to collect and identify personal and organisational names, geographic entities and subject headings. http://www.dnb.de/gnd.
The concept of linked open data is nourished by prospects of machine reasoning and automated knowledge management. Interlinking of inhomogeneous web content by use of identifiers, decidable ontology-schemata and expression of facts in simple structured phrases (RDF) seems to allow the cross-checking of information, the recomposition of knowledge and distillation of new knowledge. Digital key player adopted it (BBC, Google).
We estimated that it is to laborious to check each difference in factual assertions, especially as often a severe scientific controversy lies behind a seemingly simple difference.
In order to interoperate on linked open data and infer statements a significant amount of ontology-mapping and value-combing is necessary. The GND contained a small number of error, e.g. impossible dates. Our own linked data set contains normalised values and in some cases information is lost. For example, the year of birth "um 1225"23 or place of birth "wohl in Aquitanien"24 would require more specific vocabulary which would make inference more complicated.
The recent project, that aims to establish a historical and biographical information system, runs from 2012 to 2015 (Hockerts 2012; Jordan 2012; Hagn/Schrott 2015; Ebneth 2015). It deepens and extends the objectives mentioned in ch. 3. The Deutsche Biographie aggregates data from 15 selected renowned partners in Germany who provide content related to individuals of importance. In a first stage (2012-14) we worked with Deutsches Literaturarchiv Marbach, Bundesarchiv Koblenz/Berlin, Germanisches Nationalmuseum Nürnberg, Foto Marburg, Deutsches Rundfunkarchiv Frankfurt/Main and Deutsches Museum, München to get their individual records in selected databases linkable. 25 The second stage in 2015 includes Staatsbibliothek zu Berlin as coordinating body of "Kalliope", Deutsche Fotothek Dresden, Deutsches Filminstitut Frankfurt/Main, Landesarchiv Baden-Württemberg (Landeskundliches Informationssystem Baden-Württemberg LEO-BW26 ), as well as selected projects of four academies of sciences and humanities (Berlin-Brandenburg, North Rhine-Westphalia, Mayence and Bavaria). Partners in the second stage usually offer to align their biographical information with an authority file (GND) on their own and agree that their names should be integrated into the Deutsche Biographie.
Biographical information from our cultural heritage partners increased the Deutsche Biographie by 134.000 additional records. As of late 2014 the Deutsche Biographie offers more than 260.000 records on personalities with further biographical information and individual linking to up to 150 web resources. By the end of 2015 the number of persons in the Deutsche Biographie will rise above 500.000.
Cross linking articles by recognizing and identifying persons mentioned in the index files was a first approach to a more general extraction of information.
In addition to our newly developed faceted search we were able to offer maps with identified places of birth, death and burial as starting point. Relations between identified individuals are now additionally displayed as an ego-centered graph, utilizing the JavaScript library D3.js. The RDF-store prototype worked with a generic tool to visualise relationships (Relfinder, cf. Heim et al. 2009), but it did not offer simple graph based queries.
The accompanying RDF-store will be set up anew in OpenLink Virtuoso because of problems of speed, and scalability. The prototype developed with AKSW Leipzig (Brümmer 2011) exhibited metadata and first genealogical relations.
Current work consists of extending the ontology schema to express interpersonal relations in RDF/OWL and proceeding to express these relations: In addition to the schema of GND29 we look at Agrelon30 as a reference, but more refinement on predicates would still need to be done. Another influence was the Catalogus Professorum Model (CPM) which is expressive for statements and constraining on argument values, but it is specified on careers of university teachers and seems to be discontinued (Riechert et al. 2010a, 2010b). Further we considered the Conceptual Reference Model (CIDOC-CRM) expressed in "Erlangen OWL". It follows an "agents and events" based meta model which requires a certain overhead in order to express an interpersonal relation.
The digitisation of NDB and ADB formed a first step into "world of data". Normalisation, homogenisation, and categorisation was forced upon historical biographical accounts. Authority files offered of big help in data management and interlinking opportunities. For persons even our internal identifiers subsist because GND identifiers were not always initially available. The ontology schemata did not fit in every situation, but it can and will be technically specialised. But there are even more types of identifiers provided by authority files: works of art, not only literature, professions, organisations, fields of study and subject headings (cf. Reinert 2011b). Future work should also clarify the uncertainty of individual statements which have been normalized and respect the historical complexity of denominations (places, organisations, individuals) in an appropriate way.
The NDB received funding from the German Research Foundation (Deutsche Forschungsgemeinschaft DFG) in 2001-03, 2008-09, 2010/11 and 2012-15. 31
. http://www.ndb. badw.de#DHd_2014. Langer, Stefan, Maier, Petra, Oesterle, Jürgen (eds.) (1996). CISLEX -An Electronic Dictionary for German: Its Structure and a Lexicographic Application. In CIS- Bericht-96-97, Proceedings of COMPLEX '96, pp. 155-164. Menges, Franz, Ebneth, Bernhard (1998)