This paper argues for making a paradigm shift in publishing and using biographical dictionaries on the Web, based on Linked Data. The idea is to represent biographical data in a harmonized, semantically interoperable form, which enables 1) data enrichment by aggregating linked content from complementary, distributed, and heterogeneous data sources, as well as by reasoning, and 2) development of intelligent services using machine “understandable” data. Based on the aggregated global knowledge graph, published in a SPARQL endpoint, tooling for 1) biographical research of individual persons as well as for 2) prosopographical research on groups of people can be provided. As a demonstration of these ideas, we discuss the new in-use linked data service and semantic portal BIOGRAPHYSAMPO - Finnish Biographies on the Semantic Web that quickly attracted thousands of end users on the Web. This semantic portal is based on a knowledge graph extracted automatically from a collection of 13 100 textual biographies, written by 980 scholars. The texts are enriched with data linking to 16 external data sources and by harvesting external collection data from libraries, museums, and archives. Reasoning is used for query expansion and for discovering serendipitous relations between entities, such as persons and places.
Biographical dictionaries (Keith, 2004) may contain tens
of thousands of short biographies of historical persons of
importance. Traditionally, such dictionaries have been
published as printed book series. The Oxford Dictionary of
National Biography1 (ODNB), with more than 60 000 lives,
was first published on-line in 2004, and since then major
biographical dictionaries have opened their editions on the
Web with search engines for finding and (close) reading
biographies of interest. On-line national biographical
collections include USA’s American National Biography2,
Germany’s Neue Deutsche Biographie3, Biography Portal of
the Netherlands4, Dictionary of Swedish National
Biography5, and National Biography of Finland6 (NBF).
ODNB and other early adopters of web technology started
the paradigm shift in publishing and reading biographical
dictionaries on the Web. We call such systems 2.
generation publications. This paper argues for taking the next
step forward into 3. generation systems, i.e., to
publishing and using biographical dictionaries as Linked Data on
the Semantic Web. The goal is to serve both machine and
human readers, and support both close and distant
reading
This paper considers BIOGRAPHYSAMPO from a
publishing paradigm shift perspective, complementing our
earlier papers: In
In the following, we first present the underlying “Sampo” publishing model and series of semantic portals whose new member BIOGRAPHYSAMPO is. After this the underlying knowledge graph is presented, and the new linked data based possibilities for biographical and prosopographical research are illustrated. In conclusion, related works are discussed and contributions summarized.
7BIOGRAPHYSAMPO is available at http://biografias ampo.fi. More information and publications are available at the project homepage https://seco.cs.aalto.fi/pro jects/biografiasampo/en/.
8https://www.finlit.fi/en
The ideas of the Semantic Web (SW) and Linked Data can be applied to address the problems of (semantic) data interoperability and distributed content creation at the same time, as depicted in Fig. 1. Here the publication system is illustrated by a circle. A shared semantic ontology infrastructure is situated in the middle. It includes mutually aligned metadata and shared domain ontologies, modeled using SW standards. If content providers outside of the circle provide the system with (meta)data, it is automatically linked and enriched with each other and forms a knowledge graph. For example, if metadata about a painting created by Picasso comes from an art museum, it can be enriched (linked) with, e.g., biographies from Wikipedia and other sources, photos taken of Picasso or by him, information about his wives, books in a library describing his works of art, related exhibitions open in museums, and so on. At the same time, the contents of any organization in the portal having Picasso related material get enriched by the metadata of the new artwork entered in the system. This is a win-win “business model” for everybody to join such a system; collaboration pays off.
However, the model also creates new challenges. In
addition to enriching information also conflicting data from
different sources may be aggregated, leading to problems
of data fusion. A solution to this is to maintain provenance
metadata about the primary sources (cf., e.g.,
The model of Fig. 1 fits well with Linked Data idea of
providing data as a service and as a live SPARQL
endpoint (Heath and Bizer, 2011), on top of which
independent applications can be created on the client side without
server side concerns. We call this whole the Sampo9 model
The model has been developed and tested in a series of several practical case studies, including CultureSampo10 (2008) for cross-cultural contents, TravelSampo11 9In Finnish mythology and the epic Kalevala, ”Sampo” is a mythical artefact of indeterminate type that gives its owner richness and good fortune, an ancient metaphor of technology.
10https://seco.cs.aalto.fi/applications/ku lttuurisampo/
11https://seco.cs.aalto.fi/applications/tr avelsampo/ (2011) for tourism, BookSampo12 (2011) for fiction literature, WarSampo13 (2015) for military history, and NameSampo14 (2019) for toponomastic research of historical place names. Our experiences suggest that the Sampo model is a promising way to create useful systems that endusers like. For example, in 2018, BookSampo had ca. 2 million users and WarSampo 230 000.
In BIOGRAPHYSAMPO, the knowledge graph was
extracted from the biography collections listed in Table 1,
linked not only internally but also enriched with links to
the external data sources listed in Table 2. In addition,
data was harvested from 1) the art collection data of the
National Gallery of Finland15, 2) the National
bibliography of Finland Fennica16, 3) BookSampo semantic portal17
linked data for fiction literature
Data analysis in Digital Humanities (DH) is typically done
partly by the machine, partly by the human. In
visualizations, such as maps, timelines, and networks, the machine
presents target data in a form from which the human user
is able to make interpretations more easily. In statistics,
e.g., pie charts, line charts, and histograms are used.
Another type of tool
13https://seco.cs.aalto.fi/projects/sotasa mpo/en/
14https://seco.cs.aalto.fi/projects/nimisa mpo/
15https://www.kansallisgalleria.fi/en/avoi n-data/
16https://www.kansalliskirjasto.fi/en/news /finnish-national-bibliography-released-a s-open-data 17http://kirjasampo.fi 18http://snellman.kootutteokset.fi 19https://seco.cs.aalto.fi/ontologies/hist o/
20Hosted by the Linked Data Finland service http://ldf. fi. in order to find “interesting”, serendipitous (Aylett et al., 2012) new knowledge. Techniques such as topic modelling (Brett, 2012) fall in this category. The results also here typically need human interpretation, as statistical methods are usually unable to explain their results. In knowledge-based systems, knowledge structures can be used for this. Many of the methods and tools above are well-defined and domain independent, and there are lots software packages available for using them, such as Gephi21, R (Field et al., 2015), and various Python and JavaScript libraries. However, each of them have their own input formats and user interfaces, and need specific skills from the user. Furthermore, visualizations are crafted case by case; tools for formulating, adjusting, and comparing analysis results in some general ways would be helpful for the user.
Second generation dictionaries of biographies on the Web are used in the following traditional way: a search box or form is filled up specifying the person(s) whose biographies are searched for. Then the search button is pushed, and a list of hits is shown that can be opened for close reading by clicking. The paradigm chance of publishing biographies as linked data (third generation systems) makes it possible to build systems based on live data services, especially SPARQL endpoints. In this way, also other parties can reuse the data in their own applications. It is possible to not 21https://gephi.org only publish biographies with search interfaces, but also to incorporate ready to use tooling for DH research on top of the data service. In addition, the SPARQL endpoint makes it possible to study the data by custom designed queries in situations where the ready to use interfaces are not enough for problem solving. The SPARQL API can also be used for extracting and downloading filtered subsets of data from the endpoint in different formats (e.g., CSV) to be used in external tools, such as spreadsheets, R, or Gephi. Some of these new possibilities are illustrated below by the ready to use tools of BIOGRAPHYSAMPO.
Problem solving in DH often has two phases, as in the prosopographical research method (Verboven et al., 2007, p. 47): First, a target group of entities in the data is selected that share desired characteristics for solving the research question at hand (in the case of prosopography, a people group is selected). Second, the target group is analyzed, and possibly compared with other groups, in order to solve the research question. Using BIOGRAPHYSAMPO based on the same pattern: First, faceted search is used for filtering out a biography or a group of them for prosopography. After this, versatile ready to use tooling can be applied for reading a single biography or for analysing groups of biographies and comparing them with each other or other groups.22 Enriching the Reading Experience After finding a biography of interest, BIOGRAPHYSAMPO provides the user with an enriched reading view of the protagonist’s life by creating automatically a ”homepage” for each person, based on 1) data linking and 2) reasoning. Fig. 2 shows as an example the homepage of Eliel Saarinen (1873–1950), a prominent Finnish architect. The page contains six (6) tabs providing different biographical views of the person, here two pages based on the NBF, data at the Linked Data Finland service, a genealogical family tree and homepage by the Geni.com service, and the Finnish Wikipedia article. The entry is linked to seven (7) external data sources on the 22A short video is available on the Web illustrating the ideas of BIOGRAPHYSAMPO: https://vimeo.com/328419960. web. On the right, recommendation links to related biographies are given, e.g., to similar biographies based on their linguistic content. On the top of the page, there are five (5) tabs providing data-analytic views of Saarinen.
Network Analysis For example, Fig. 3 presents his egocentric network based on the links between the bios in the NBF, with a coloring scheme indicating persons of different types. The depth and other parameters of the network can be controlled by the widgets on the left. In Fig. 4, another tab visualizes the international events of Saarinen’s life on a map and on four timelines of different event types (personal life, career, artistic or scientific creations, and accolades) for a spatiotemporal analysis.
pography, BIOGRAPHYSAMPO employs faceted search for filtering out not only individual persons but also groups of them sharing some properties, such as profession, place of birth, place of education, working organization, etc. Once the target group has been selected, various generic data-analytic tools and visualizations can be applied to the group: 1) Statistical tools include histograms showing various numeric value distributions of the biographees, e.g., their ages, number of spouses and children, and pie charts visualizing proportional distributions of professions, societal domains, and working organizations. 2) Event maps show how different events (personal life events, career events, artistic and scientific creation events, and accolades) participated in by the biographees are distributed on maps. 3) Life charts summarize the lives of persons from a transitional perspective as blue-red arrows from the birth places (blue end) to the places of death (red end).
These tools and visualization can be applied not only to one target group but also to two parallel groups in order to compare them. For example, Fig. 5 compares the generals and admirals of the Grand Duchy of Finland (1809–1917) (on the left) with the clergy (1800–1920) (on the right). With a few selections from the facets the user can see that, for some reason, quite a few officers moved the to south to die while the Lutheran ministers stayed more in Finland. The arrows are interactive. For example, by clicking on the peculiar upper arrow to the east, one can find out that this arrow was due to general Gustaf A. Silfverhjelm’s (1799– 1864) biography, where one can learn that he become a chief cartographer in western Siberia where he died. Searching for Historical Places BIOGRAPHYSAMPO also provides the user with a map search view that projects the places in which the ca. 100 000 biographical events extracted from the biographies are projected on the places where they occurred. The maps in this view are not only contemporary ones but also historical maps served by a separate historical ontology and map service Hipla.fi23. Many important events of Finnish history took place in the eastern parts of the country that was annexed to the Soviet Union after the Second World War. Old Finnish places there may have been destroyed, placenames been changed, and names are now written in Russian. Using semi-transparent historical maps on top of contemporary maps solves the problem by giving a better historical context for the events.
knowledge discovery in Linked Data, an application
perspective for finding ”interesting/serendipitous” (Aylett et
al., 2012) connections in the biographical knowledge graph
was created. This application idea is related to relational
search
Aside the business of publishing biographical dictionaries
in print and on the Web, representing and analyzing
biographical data has grown into a new research and
application field. In 2015, the first Biographical Data in
Digital World workshop BD2015 was held presenting several
works on studying and analyzing biographies as data (ter
Braake et al., 2015),
These lines of research are related to ours as they are based on the idea of extracting semantic structures from the largely unstructured biographical text collections, and on using the data for DH research in biography and prosopography. In addition and in contrast to the related works, BIOGRAPHYSAMPO employs the “Sampo model” where the data is enriched through a shared content infrastructure by related external heterogeneous datasets, here, e.g., collection databases of museums, libraries, and archives, a 24http://www.biographynet.nl 25http://www.sixdegreesoffrancisbacon.com critical edition, genealogical data, and various biographical data sources and semantic portals online. Another difference is that in our work, a main goal has been to develop and provide versatile DH tooling for end-users on top of a Linked Data SPARQL endpoint.
This paper presented and demonstrated the vision of a paradigm shift in publishing biography collections on the Semantic Web. The vision has also been operationalized and implemented as the semantic portal BIOGRAPHYSAMPO now in use on the Web by thousands of users. The biographical data of the portal was extracted and aggregated automatically by the computer and has not been fully validated by human experts, which would be impossible due to the amount and complexity of the big data. This is a typical situation in DH research, and calls for using more source criticism when interpreting the analyses than when dealing with human curated datasets. The quality and completeness of the BIOGRAPHYSAMPO data has not yet been analyzed formally, but our informal tests suggest that the results are very useful even if errors are also encountered. This is the price to be paid for advanced end-user services and distant reading on distributed heterogeneous biographical data. Acknowledgements This research was part of the Severi project26, funded mainly by Business Finland. Thanks to CSC – IT Center for Science, Finland, for computational server resources for the data service and applications.
R. S. Aylett, D. S. Bental, R. Stewart, J. Forth, and G.Wiggins. 2012. Supporting serendipitous discovery. 26http://seco.cs.aalto.fi/projects/severi