Converted from an OASIS Open Document
This paper presents ongoing foundational theoretical and practical work on the application of ontology-based modeling to represent and visualize the complexity of knowledge disseminated in historical narratives. In short, the new approach combines modeling informed by philosophical ontology and philosophy of history with semiotically founded visualization of historical processes in order to support historical understanding.
The following quote from Munslow (2007) lends itself as an appealing summary of the character of historical narratives: “In writing a history for the past we create a semiotic representation that encompasses reference
to it, an explanation
of it and a meaning
for it.” What role could information visualization tools play in this context? As Champagne (2016) remarks, “[h]istorians occasionally use timelines, but many seem to regard such signs merely as ways of visually summarizing results that are presumably better expressed in prose.” He challenges this view and argues that timelines could support the historian in gaining novel historical insights. The main cognitive funtion of timelines is according to Champagne (2016: 40) the “logical conjunction by visual juxtaposition”. Furthermore there is also the potential of abductive reasoning: “Timelines, however, are more likely to surprise us, by showing us past events that we would have never otherwise considered chunking. Hence, in addition to historical scholarship expressed in regular prose, consulting diagrammatic signs can foster the discovery of patterns essential to a fuller understanding of the past” (Champagne, 2016: 40). This is especially more likely if there are synchronoptic timelines showing historical events of different categories—i.e. not only political events, but also economic or cultural events etc. (That is the approach conducted by Peters and Peters (1952) in their
Synchronoptische Weltgeschichte.) For example, such parallel timelines could possibly be used as a tool to support periodization (see possible use case reported by Luyt (2015)).
A “visual historiography” (Roegiers and Truyen, 2008) can quite easily be done via the temporal, spatial, and thematic context of information about historical events, but without explicitly stated relations between events it is questionable how useful that could be in supporting historical research. The big advantage of that approach is of course, “that one is able to represent the complexity of a historical subject, without having to fill out the gaps, or having to choose between different interpretations, but using an [information integration] architecture that places the subject in its context(s)” (Roegiers and Truyen (2008: 70) as cited in Sabharwal (2015: 57)). The problem with such a ‘visual historiography’ is that it cannot support visual contextualization done by the historian during conceptualizing complex interrelations of historical events—including not only temporal, spatial, and thematic relations, but also causal relations (incl. the motivation and roles of historical actors involved in the events), mereological, and constitutive relations of complex (e.g. composite) events.
Anyhow, explicit modeling of event structure and relations is necessary because without a more fine-grained representation of the structure and interrelations of events visualization tools are indeed limited to bare juxtaposition. Digital history demands information visualizations beyond simple timelines (see for example Drucker and Nowviskie, 2003). Diagrammatic approaches for multiperspectival analysis and synopsis of historical sources are needed. There are rarely technical implementations and just a few theoretical approaches to the development of such tools for multiperspectival exploration of historical sources (cf. Shaw, 2008: 90).
intrinsic relations to other events and locating it in its historical context” (Walsh (1951 S. 59) as cited in Shaw (2010 S. 11)). An adequate modeling of colligatory concepts and the relations between the concepts—the colligatory relations (Shaw, 2008)—is the precondition for “semantic tools” (Shaw, 2013) based on such an explicit representation of the past.
The digital edition (Behrendt et al., 2010) of Peter’s (already mentioned)
Synchronoptische Weltgeschichte (Peters and Peters, 1952) visualizes parallel timelines showing historical events from different categories (political events, economic events, cultural events, etc.). However, it does not show the inner structure of complex events or processes and does not provide typed relations between events (see Shneiderman, 1996). Interestingly the tool provides visual contextualization of events based on the knowledge organization in its event database (see Fig. 1): related events are retrieved based on their common index terms.
The older tool
SemTime (Jensen, 2003) provides a solution to these typical shortcomings of timeline visualizations by introducing Semantic Timelines to visualize complex timelines with sub-timelines and different types of relations between historical events. Newer projects concentrate not only on the granularity of events but also on the the details of biographies, i.e. the modeling and visualization of the roles of (historical) persons in events (Trame et al., 2013; Hyvönen et al., 2018).
The VICODI (Visual Contextualisation of Digital Content) project (Nagypál et al., 2005) used semantic web ontologies as basis for visual contextualization. Based on the top-level ontology DOLCE (Gangemi et al., 2002) the SHO (Spatial History Ontology) was developed by Grossner (2010) to overcome the shortcomings of exisiting ontologies in the modeling of spatial information in event ontologies. HERO (Historical Event Representation Ontology) is also founded on DOLCE and focuses on the modeling of different types of roles (thematic, social, and also perspectival) (Goy et al., 2018). DOLCE and DOLCE-DnS Ultralite (DUL) respectively contains the Descriptions and Situations (DnS) Ontology Design Pattern (ODP) (Gangemi and Mika, 2003). DnS allows the modeling of different perspectives on entities. A CRM (Doerr, 2003) based alternative to model perspectives or interpretations is the MIDM (Multiple Interpretation Data Model) (Ruymbeke et al., 2017). A much simpler modeling approach for different perspectives is SEM (Simple Event Model) (Hage et al., 2011).
A crucial requirement for the approach presented here is the representation of different perspectives on historical events. Perspectival explanation or “synoptic judgement” (Mink, 1987) is a main task of the historian (cf. Levy, 2001: 70). I have chosen DUL as top-level ontology for the modeling examples described in the following paragraphs because of its constructivist design principles and especially because its DnS ODP fits very well to our requirement of modeling colligations and the different perspectives or interpretations of historical events. Thus, a Description represents the conceptual relations which were grasped by the historian in a synoptic judgement.
Fig. 2 shows a screenshot from an experimental tool that draws diagrams of causal narratives. The example is from Theda Skocpol’s
States and Social Revolutions: A Comparative Analysis of France, Russia, and China. Skocpol (1979) describes the historical process which led to the French revolution in narrative form. In order to visualize the historical process consisting of three sub-processes, the mereological and causal relations had to be represented in a knowledge base according to the reconstruction of Skocpol’s narrative done by Mahoney (1999).
Skocpol (1979) combines macrosocial and idiographic historical research (cf. Mahoney, 1999: 1189). Fewer idiographic detail is represented in the example from the CEWS (Conflict Early Warning Systems) project (Schmalberger and Alker, 2001a) (see Fig. 3). CEWS focuses on the phase sequences (escalation and de-escalation) in conflict processes. The CEWS Explorer was developed as a tool to visualize and compare conflict phase sequences and different perspectives on them as described in causal narratives about conflicts. In my talk I argue that a remake of this approach can benefit from an ontology-based representation of conflict phases and different perspectives on phases and causes for change of phases as seen from different conflict parties or other actors involved in the conflict.
According to Peirce’s diagrammatic reasoning approach a diagram should be constructed under the rules of a “system of representation” (CP 4.418). The ontology-based knowledge representation of historical events provides as well a framework for the construction of such representation systems. Similar to the so-called two-level theory of social revolutions in the first example (see diagrammatic representation of the theory in Goertz and Mahoney (2005: 509)) the so-called visual grammar of possible conflict phase sequences in Fig. 4 is a representation system that pretends all possible sequences of different types of conflict phases within conflict episodes. The system is used as “system of diagrammatization” (NEM IV:318) in order to construct diagrams for specific conflict trajectories.
As there is less granularity, i.e. no complex composite events, a simpler ODP can be used to represent conflict phase transitions. Fig. 5 shows an exemplary phase transition modeled with the ODP Transition
There are two feasible use cases for the presented modeling and visualization approach—followed by a more demanding one:
The added value for the diagrammatic communication of research findings originates in the explicit representation of historical knowledge: A complex historical process can be visualized for better communication of research results on the base of the explicitly represented entities and their relations. As Lange (2013: 46) notes in his textbook on comparative-historical methods, it is recommended to use “diagrams to represent clearly the argument of causal narration, to make the causal claim more explicit”.
Supported by a ‘system of diagrammatization’, a (public) historian is enabled to represent and visualize the essential event relations of a “basic story” (cf. Perfetti et al., 1995: 2)—excluding more granular expert knowledge about the historical process in focus. In knowledge visualization projects for public history additional diagram types could be used in order to provide the “collateral knowledge” (Hoffmann, 2005) necessary for the public audience (non-historians) to better understand the historical events—e.g. via concept maps or knowledge mapping in general (Davies, 2011).