Historical Knowledge Modelling and Analysis through Ontologies and

Timeline Extraction Operators: Application to Computing Heritage

Christophe Ponsard

, Aur

elien Masson

and Ward Desmet

CETIC Research Centre, Charleroi, Belgium

ESEO Engineering School, Angers, France

NAM-IP Computer Museum, Namur, Belgium

Keywords:

Domain Speciﬁc Modelling, Ontology, Historical Knowledge, Visualization, Timeline, Case Study.

Abstract:

Cultural heritage as human science has a long tradition of text-based reporting and analysis. This domain

has a very rich semantic structure, especially to relate many different types of entities anchored in some time

period with more or less strong temporal inter-dependencies. Various modelling approaches, largely based on

ontologies, have been proposed to capture and structure this kind of knowledge. In this paper, we are inter-

ested in easing the analysis capabilities on historical knowledge using the timeline as central concept that can

be extracted and manipulated in various ways through speciﬁc operators sharing some similarities with multi-

dimensional analysis in business intelligence. We propose zooming on a speciﬁc aggregates, pivoting from a

timeline to another one or drilling-across to compare different timelines. Our work is illustrated on a concrete

implementation targeting the Computing Heritage of the micro-computer period, including machines, operat-

ing systems, companies, people and applications. The information is extracted from a museum information

system combined with DBPedia. We also developed a speciﬁc visualisation tool under the form of a mobile

application which can also be used as museum guide.

1 INTRODUCTION

As deﬁned by the SAGE dictionary (Thorpe and Holt,

2007): “Historical analysis is a method of the exam-

ination of evidence in coming to an understanding of

the past. It is particularly applied to evidence con-

tained in documents, although it can be applied to

all artefacts. The historian is, ﬁrst, seeking to gain

some certainty as to the facts of the past. Establish-

ing the facts also gives the researcher a chronology.

The second task is to seek to establish cause and ef-

fect between those facts in order to understand why

things happened. It is important to remember that

while the past is the immensity of everything that has

happened, history is what we know of the past”. The

way the past is presented is usually based on a nar-

rative because “Human beings are story tellers who

exist ontologically in a universe of narrative making”

(Ankersmit, 2005).

The above references highlight fundamental as-

pects of historical analysis and reporting:

• history is what we know about the past,

• its analysis is based on the identiﬁcation of the

chronology and casual relations across facts, and

• its reporting is done through some form of narra-

tive referring to an ontology.

The development of computer technologies and net-

works has led to the emergence of digital history

which goes far beyond the pure digital formatting of

content. It also enables new forms of historical anal-

ysis and visualisations (Burton, 2005), especially “vi-

sual historiography” which integrates non-textual ma-

terials into historical representations and hypertexts

with social, spatial, and chronological perspectives

(Roegiers and Truyen, 2006).

This paper focuses on a such visualisations

through a common form of visual representation for

time-based narrative: the timeline. Beyond the mere

chronological representation of events in chronologi-

cal order, the way it is build and displayed can exploit

a rich set of relationships such as mereology (is part

of), causality (is cause/consequence) and thematic as-

pects (tags). This requires conceptualisation and the

explicit modelling of event structure and relationships

through an adequate meta-model or ontology. Provid-

ing ﬂexible and dynamic timeline visualisation means

is useful both for:

• understanding history by engaging with the model

302

Ponsard, C., Masson, A. and Desmet, W.

Historical Knowledge Modelling and Analysis through Ontologies and Timeline Extraction Operators: Application to Computing Heritage.

DOI: 10.5220/0010900400003119

In Proceedings of the 10th International Conference on Model-Driven Engineering and Software Development (MODELSWARD 2022), pages 302-309

ISBN: 978-989-758-550-0; ISSN: 2184-4348

to produce explanations for some events by trac-

ing its connections with other events and the his-

torical context (Walsh, 1951), and

• providing speciﬁc views and explanations to the

general public in the context of an exhibition to

stress the importance/inﬂuence of speciﬁc events

or of a more global context, e.g. through parallel

related timelines.

The use case which triggered this work is an exhi-

bition about the micro-computer revolution from the

1970’s to the 1990’s which is technologically an-

chored in the development of microprocessors and

graphical user interfaces, and socially in the hacker

counter-culture. This work involved many timeline

visualisation both in the preparation phase and as vi-

sualisation support for the exhibition. In order to

go beyond static paper posters, we also designed an

application providing a dynamic form of navigation

through timelines. This work has led us to reﬁne and

formalise our approach with the aim to deﬁne a more

general framework and tooling. The result could be

applied in a wider context: in a computer museum

but also in other kinds of museums or organisations

involved in historical studies. We report here about

our current progress and the following contributions:

• the design of the underlying ontology, based on a

comparative survey,

• the extraction of interesting timelines from the re-

sulting knowledge base, and

• the illustration of speciﬁc visualisations though

our application.

This paper structure follows the above contributions.

Our micro-computer case study is used as running ex-

ample. Sections 2, 3 and 4 will respectively cover on-

tology design, timeline extraction and visualisation.

Section 5 will discuss our results. Finally, Section 6

will conclude and give some perspectives.

2 ONTOLOGIES FOR

HISTORICAL KNOWLEDGE

Many ontologies and meta-models have been pub-

lished and provide solid grounds to build a knowledge

base. This section reviews the main representatives

and compares them against key concepts for digital

history processing.

2.1 Simple Event Model (SEM)

Events have become central elements in the represen-

tation of data from domains such as history, cultural

heritage, multimedia and geography.

The Simple Event Model (SEM) is an ontology

created to model events in various domains such as

history, cultural heritage, multimedia and geography,

without making any assumptions about the domain-

speciﬁc vocabularies used (W3C, 2009)(van Hage

et al., 2011). It is designed with a minimum of se-

mantic commitment to guarantee maximal interoper-

ability. SEM is structured around four core concepts

depicted in Figure 1:

• Event is the central concept. It has a loose def-

inition (e.g. possibly ﬁctional) and can originate

from different sources. To cope with this, events

may have bounded roles, related facts may have

time bounded validity and an authoritative source

may be associated to views.

• Actor: who or what participated in an Event.

• Place: where the an Event took place.

• Time: when the Event took place.

In this case study statement “The UK company Sin-

clair Radionics Ltd was founded by Clive Sinclair in

1961”: the Event is the foundation, the Actor is Clive

Sinclair, the Time is 1961 and Place is UK.

Figure 1: Simple Event Model.

Historical Knowledge Modelling and Analysis through Ontologies and Timeline Extraction Operators: Application to Computing Heritage

303

Figure 2: DOLCE Foundational Ontology.

2.2 DOLCE and Extensions

DOLCE (Descriptive Ontology for Linguistic and

Cognitive Engineering) is a foundational ontology

(FO) developed originally in the EU WonderWeb

project (Masolo et al., 2003). FOs are domain-

independent axiomatic theories that provide solid

grounds to capture and reason about knowledge. Fig-

ure 2 depicts the meta-model using an UML class di-

agram. It is structured using the dual concepts of En-

durant (including Objects or Substances) and Perdu-

rant (including Events, States, or Processes) which

are linked by the relation of participation:

• Endurants: localized in space, they get their tem-

poral location from Perdurants they participate in.

• Perdurants: localized in time, they get their spa-

tial location from Endurants participating in them.

• Additionally, speciﬁc qualities can characterise

either Endurants (as Physical or Abstract Quali-

ties) or Events (as Temporal Qualities). Each kind

of Quality is associated to a Quality Space repre-

senting its range of possible values.

Figure 3: Spatial History Ontology.

Examples of Endurants in our exhibition are

micro-computers (e.g. IBM-PC, ZX81) and people

like computer scientists or entrepreneurs (e.g. Bill

Gates, Steve Jobs). However, the later have can ini-

tiate events while the former are pure physical ob-

jects. This ﬁner-grained level is supported by ex-

tended ontologies such as Spatial History Ontology

(SHO) (Grossner, 2010) shown in Figure 3.

2.3 Constructed Past Theory (CPT)

Constructed Past Theory is an epistemological theory

about how we come to know things that happened or

existed in the past (Thibodeau, 2019). It is more elab-

orated in making the distinction between constructed

and target past. Is it speciﬁed as UML class diagram

partly depicted in Figure 4.

Figure 4: Constructed Past Theory (UML meta-model).

MODELSWARD 2022 - 10th International Conference on Model-Driven Engineering and Software Development

304

Figure 5: Main Structure of DBpedia.

It covers the SEM concepts of Event and En-

tity, including active. However, space and time

are captured as attributes and not as ﬁrst class con-

cepts as in SEM and DOLCE. More complex events

can be captured through processes which is also

present in DOLCE (aggregated Perdurant) and SEM

(subevents).

2.4 DBPedia Ontology

The previous meta-models are interesting but in order

to be practical, data must be available. A rich form of

data is provided by wikipedia. Structured information

(e.g. infoboxes) is collected by DBPedia (Bizer et al.,

2009)(Cvjetkovic et al., 2013)(Lehmann et al., 2015).

It is organised into a dedicated semantic web ontology

depicted in Figure 5. It shows the presence of the key

concepts discussed before: general Thing and a more

specialised form of Agent. The notion of Space is also

explicit.

However, the notion of Event is not a ﬁrst class

concept: time aspects are only attached to speciﬁc

concepts as attributes, e.g. birth and death date of a

Person, release date or a Work. Consequently, build-

ing timelines requires to perform complex queries to

gather the relevant information possibly across differ-

ent concepts. DBPedia can be queried using SparQL

(SPARQL Protocol and RDF Query Language), a

standardised semantic query language for databases

able to retrieve and manipulate data stored in Re-

source Description Framework (RDF) format (OMG,

2008).

Listing 1: SparQL query over computer companies.

SELECT DISTINCT ? da t e ?name

WHERE {

? company f o a f : name ? name .

? company dbo : i n d u s t r y d b r : C o m p u t e r ha r d w a r e .

? company dbo : fou nde dBy ? f o u n d e r .

? company dbo : f o u n d i n g Y e a r ? da t e .

FILTER ( ? d a t e <= ” 1 9 9 0 0 1 0 1 ” ˆ ˆ xsd : d a t e )

}

ORDER BY ASC( ? d a t e )

LIMIT 100

Listing 1 shows how to retrieve companies active

in computer hardware founded before 1990. The for-

mulation of the query requires careful design as data

can be quite heterogeneous. This query yields the re-

sult partially depicted in Listing 2 which represents a

timeline. Note that extra attributes retrieved by the re-

lated query are not shown but can be used e.g. founder

and other company attributes. Note the interesting du-

plicate: HP Inc. and Hewlett-Packard. It can be ex-

plained by the split of this company in 2015. Such

evolution is not yet captured by our current frame-

work.

Listing 2: SparQL query over computer companies.

d a t e name

1911 I n t e r n a t i o n a l B u s i n e s s Mach i nes C o r p o r a t i o n

1939 HP In c .

1939 Hewlett − P a c k a r d Company

1958 Commodore I n t e r n a t i o n a l C o r p o r a t i o n

1959 K o n t r o n S\&T AG

1960 Augmentation Re s e a r c h

. . .

Historical Knowledge Modelling and Analysis through Ontologies and Timeline Extraction Operators: Application to Computing Heritage

305

3 TIMELINE PROCESSING

This section details how our timelines are represented

and manipulated. First, we present our meta-model

design which is strongly inspired from the ontologies

reviewed in Section 2. Second, we identify the main

timeline extraction and navigation operations in order

to provide interesting ways to walk through a domain

using the time dimension as main representation but

also by exploiting various ﬁlters and interconnections

present in our meta-model.

3.1 Meta-model Design

The survey in the previous section illustrated that the

identiﬁed ontologies share key concepts of event, ac-

tive/passive objects and links with space and time in-

formation but with some differences on how to gener-

alise or specialise them. We did not select a speciﬁc

ontology literally but produced a customised meta-

model strongly aligned with them, only keeping com-

mon and relevant concepts ﬁtting our need to reason

on timelines. Figure 6 depicts its current version.

Figure 6: Current Timeline-oriented Meta-model.

Our approach was to combine the strong theoret-

ical background based on DOLCE/SHO while keep-

ing it as simple as possible like SEM while enabling

to rely on information extracted from DBPedia. Our

goal is to build a high quality knowledge base by

combining ﬁrst class data sources like museum in-

ventory and implementing a cleaning and validation

process by relying on open data. We explicitly mod-

elled the Timeline as ﬁrst class concept because it is

central to our time-oriented approach while Space is

captured using a Location concept. We refer to Per-

durant (Events) and Endurant (either passive: Ob-

jects or active: Actors) using stereotypes. We sup-

port aggregation for all concepts with possible sub-

timelines (e.g. speciﬁc phases of micro-computer de-

velopment), sub-locations (e.g. a speciﬁc country,

town), groups for actors (to capture the notion of com-

pany, department, e.g. XEROX PARC) and objects

sub-parts (e.g. computer monitor, central unit, key-

board). An Object can be of physical but also logical

nature, e.g. software in our case.

3.2 Timeline Extraction and Navigation

Assuming we have achieved data collection and vali-

dation according to our meta-model, we can consider

the user experience going through the knowledge base

using timelines. It is interesting to explore what kind

of timeline can be extracted and how they can be con-

nected together. We could identify different meaning-

ful timelines that collect related events by focusing

on:

• actor(s) at different granularity levels: it can be

the life of a person, a group or a company, possi-

bly with a focus on common event characteristics,

• object(s), also at different granularity levels: it

could relate to the precise history of a speciﬁc ob-

ject (e.g. the design of the LISA computer) but

also of a family of objects according to speciﬁc

criteria, e.g. micro-computer of a speciﬁc period,

manufacturer, using a speciﬁc CPU,...

• temporal, spatial or thematic contexts, respec-

tively through speciﬁc Event (dates), Location or

Tag characteristics. Different granularity levels

can also be considered, e.g. to reﬂect the com-

puter history related to micro-computer in France

from 1970 to 1985. This can come as additional

ﬁlter for the previous types of timelines.

Moreover timelines can be enriched by following

speciﬁc dependencies such as causality links or ex-

pressed by explicit references to secondary enti-

ties. However, this can result in losing focus so we

favoured a more dynamic mechanism through the use

of different navigation operators allowing the user to

easily “switch”between timelines according to spe-

ciﬁc aspects he would like to investigate. This shares

some similarities with OLAP operators in multidi-

mensional analysis, although we only focus on time

here (Salley and Codd, 1998).

Figure 7 reﬂects this idea by showing 6 timelines

MODELSWARD 2022 - 10th International Conference on Model-Driven Engineering and Software Development

306

with different scopes (Person, Company, Operating

System,...) with intersecting points through speciﬁc

events which can be used as pivot to change perspec-

tive. Some possible operators are:

• event pivoting between related entities or features:

e.g. from Amiga 500 computer to Commodore

Company or the 68k CPU or GUI timeline.

• time zoom in/out based on a deﬁned period, e.g.

the micro-computer history can be divided in

early, golden age and standardisation periods.

• actor zoom in/out, from person level to company.

• object zoom in/out, e.g. down to version/variant

level and up to product family level.

• relations inclusion, possibly iterative and with

closure, e.g. to follow casual relation to look for

causes/consequences related to some events.

• combining multiple timelines together either

merged or keeping them separated with an ad-

equate visualisation (temporal alignment, shared

events, speciﬁc relations...)

Figure 7: Navigation through Pivoting.

4 IMPLEMENTATION

Our current implementation is a monolithic proto-

type mobile application composed of two main com-

ponents: a database and a visualisation module.

The database implements our meta-model using a

lightweight SQLite relational database (Hipp, 2000)

fed by local computer inventory and DBPedia. The

visualisation relies on React Native (Facebook, 2015)

which provides a nice interface for a museum guide.

It can later evolve towards a client/server application

and is already available in Open Source (NAM-IP,

2021). Figure 8 shows a typical usage scenario with a

pivoting between two timelines, from left to right:

1. a timeline showing some computers of the second

phase (1977-1990),

2. a navigation to the Amiga 500 computer launched

in 1987 displaying features such as the manu-

facturer, location, Operating System (OS), all of

which are possible entry points to other timelines,

3. as the user decided to pivot to OS, the OS timeline

is shown with a focus on the Amiga OS.

The user is also able to conﬁgure its timeline by se-

lecting a speciﬁc period or the whole history, and by

enabling or disabling different thematic features such

as machines (MICRO), operating systems (OS), pro-

cessors (CPU), user interface (IHM) or applications

(APP). At this point, all events are merged and only

the start date is considered so possible overlaps are

not reﬂected. In order to have a better view, especially

on smaller displays, a compact mode not showing pic-

ture nor ﬁrst lines of description is available.

Figure 8: Navigation Scenario through our Mobile Guide.

Historical Knowledge Modelling and Analysis through Ontologies and Timeline Extraction Operators: Application to Computing Heritage

307

5 DISCUSSION OVER RELATED

WORK

The use of timelines in visual historigraphy has been

discussed in the introduction. From a research per-

spective, the importance of tools providing visual

contextualization of events is well-known (Shneider-

man, 1996). We focus here on software development

sharing similarities with our approach.

ChronoZoom is an open source data visualiza-

tion tool that supports zooming through time to ex-

plore timelines across the whole history of the uni-

verse down to present history of humanity (Walter

et al., 2013). It is fully web-based and has an ef-

ﬁcient Cloud-based hosting providing efﬁcient “inﬁ-

nite” zooming capabilities. It allows the user to cre-

ate or customize timelines and to plot various time

series data next to each other for comparison. Our

approach is less ambitious as we only focus on spe-

ciﬁc thematic typically the domain of a museum. Al-

though we focused on a speciﬁc exhibition, our ap-

proach is general and inspired by major ontological

frameworks but also practical semantic web technolo-

gies, which opens the way to grab information from

Wikipedia contributions. As ChronoZoom, we target

both education and research but our current front-end

is more focused on education. At the technological

level, our solution can grab data from a locally em-

bedded database but possibly also from a web API

although not optimised for large scale use.

Different mobile applications have been devel-

oped more speciﬁcally for supporting computer his-

tory in general or in the context of a museum collec-

tion. They rely on the notion of timeline more or less

explicitly. The Nexon Computer Museum, the ﬁrst

Korean Computer Museum with approximately 7,000

artifacts has a complete application to browse through

the 400 artefacts on display and relying on a chrono-

logical ordering (Nexon, 2013). It gives details but

few navigation operations. A computer history time-

line developed by a company specialised in guides is

also available for Android (Every Time Apps Studio,

2018). It is close to a poster timeline with however

a practical slider for quick access. Different websites

also publish computer timeline, including the famous

Computer History Museum (CHM, 2021). It recaps

the main events with a zoom by decades and years. It

also proposes classiﬁcation in different themes as we

do, and additionally a search engine.

Gathering and qualifying data is a difﬁcult task

and can beneﬁt from the semantic web. Our current

approach relies on DBPedia (Bizer et al., 2009) but

other knowledge graphs could be used such as the

one provided by Google (Google, 2012), although

it is itself largely based on Wikipedia. A difﬁculty

is to achieve a mapping towards on our meta-model

given the diversity of ﬁelds which are not uniformly

used. Moreover the content is largely pure text, which

means many interesting information cannot be auto-

matically extracted with semantic search engines. A

solution to this is the use of a semantic extension

enabling to make semantic content explicit and thus

make it visible to DBPedia (Vrandecic and Krotzsch,

2005). Another/complementary approach is to rely

on natural language processing to extract such knowl-

edge provided some form of reliability can be en-

forced. In our approach, we introduced a basic form

of semantic processing by detecting synonyms in our

text description, e.g. ”First Macintosh” or ”Macintosh

128K” or ”Apple Macintosh” which can be extracted

from DBPedia and manually enriched as required.

Specialised work on the formulation and recognition

of temporal patterns of events in English could also

be considered (Saygi et al., 2018).

6 CONCLUSION & NEXT STEPS

In this paper, we presented our current progress in

developing a framework for modelling and analysing

historical knowledge by relying on timeline extrac-

tion, navigation and visualisation techniques. Al-

though anchored in a speciﬁc case study of computer

heritage, we took care of setting solid foundations

based on a survey of relevant ontologies. We could

successfully implement a prototype application based

on a designed meta-model populated by a validated

dataset mixing in-house and DBPedia information.

The application, currently under validation in our mu-

seum, is quite appreciated.

Our framework is already available in Open

Source on Github (NAM-IP, 2021). In the future,

we plan to extend it in different directions: introduce

more pivot points in the user interface, support par-

allel visualisation of timelines, provide location ﬁl-

tering and new categories for more cultural or politi-

cal context and references. We also plan to elaborate

a web-service API with a robust and scalable graph

database as backend. At longer term, a collaborative

web client with semantic editing capabilities will be

considered.

ACKNOWLEDGEMENTS

We thanks the volunteers and anonymous visitors of

the NAM-IP computer museum for their feedback on

our current prototype.

MODELSWARD 2022 - 10th International Conference on Model-Driven Engineering and Software Development

308

REFERENCES

Ankersmit, F. (2005). Sublime Historical Experience.

Cultural memory in the present. Stanford University

Press.

Bizer, C. et al. (2009). DBpedia - A crystallization point for

the Web of Data. Journal of Web Semantics, 7(3):154–

165. The Web of Data.

Burton, O. V. (2005). American Digital History. Social

Science Computer Review, 23(2):206–220.

CHM (2021). Timeline of computer history. https://www.

computerhistory.org/timeline.

Cvjetkovic, V., Djokic Petrovic, M., and Arsic, B. (2013).

Wikipedia Browsing With DBpedia. In The 2nd Elec-

tronic International Interdisciplinary Conference.

Every Time Apps Studio (2018). History of Computer.

Google Play Store.

Facebook (2015). React native. https://reactnative.dev.

Google (2012). Knowledge Graph. https://developers.

google.com/knowledge-graph.

Grossner, K. (2010). Representing Historical Knowledge in

Geographic Information Systems. PhD thesis.

Hipp, D. R. (2000). Sqlite. https://www.sqlite.org.

Lehmann, J. et al. (2015). Dbpedia - a large-scale, multi-

lingual knowledge base extracted from wikipedia. Se-

mantic Web, 6(2):167–195.

Masolo, C., Borgo, S., Gangemi, A., Guarino, N., and

Oltramari, A. (2003). Ontology Library D18. Won-

derWeb project.

NAM-IP (2021). Mobile museum guide. https://github.

com/NAMIP-Computer-Museum/guideApp.

Nexon (2013). Mobile application of the computer

museum. https://www.nexoncomputermuseum.org/

english.

OMG (2008). SPARQL Query Language for RDF. https:

//www.w3.org/TR/rdf-sparql-query.

Roegiers, S. and Truyen, F. (2006). History is 3D: present-

ing a framework for meaningful historical representa-

tions in new media. New heritage: beyond verisimili-

tude, pages 49–57.

Salley, C. and Codd, E. F. (1998). Providing OLAP to User-

Analysts: An IT Mandate.

Saygi, G., Blaise, J.-Y., and Dudek, I. (2018). Temporal

hints in the cultural heritage discourse: what can an

ontology of time as it is worded reveal? Revue des

Nouvelles Technologies de l’Information.

Shneiderman, B. (1996). The eyes have it: A task by data

type taxonomy for information visualizations. IEEE

Computer Society.

Thibodeau, K. (2019). The construction of the past: To-

wards a theory for knowing the past. Information, 10.

Thorpe, R. and Holt, R. (2007). The SAGE Dictionary of

Qualitative Management Research. SAGE Publica-

tions.

van Hage, W. R., Malais

e, V., Segers, R., Hollink, L., and

Schreiber, G. (2011). Design and use of the Sim-

ple Event Model (SEM). Journal of Web Semantics,

9(2):128–136.

Vrandecic, D. and Krotzsch, M. (2005). Semantic wiki.

https://www.semantic-mediawiki.org.

W3C (2009). ”simple event model”. https://semanticweb.

cs.vu.nl/2009/11/sem/.

Walsh, W. H. (1951). An Introduction to Philosophy of His-

tory.

Walter, R. L., Berezin, S., and Teredesai, A. (2013).

Chronozoom: Travel through time for education, ex-

ploration, and information technology research. As-

sociation for Computing Machinery.

Historical Knowledge Modelling and Analysis through Ontologies and Timeline Extraction Operators: Application to Computing Heritage

309