Extraction of Spatio-Temporal data about

Historical events from text documents

Case Study: German-Herero war of resistance 1904

23 July 2018

Faculty of Environmental Sciences, Department of Geosciences, Chair of Geoinformatics

Susanna Ambondo Abraham Alumni: MSc in Cartography

Stephan Maes; Lars Bernard

TU Dresden, Chair of Geoinformatics

Content • Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions & Recommendations

23 July 2018 2

• History describes geography in the past.

• Space-time geography –location & time of occurrence.

• Free access to historical digital archives

• Transform text documents into GIS representations – NL & IE techniques

Motivation

• Better understanding of IE for historical spatio-temporal data.

• Better understanding of events on the German - Herero war of

resistance. 3

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Case Study Background

• 1880s German Settlers arrived in SWA.

• Spread across the country

• Early 1900’s the resistance struggle

began.

• Hereros revolted in 1904.

• Germany responded by sending

approx. 15000 troops under General

Von Trotha.

• Battle of Hamakari, 11 August 1904 –

Hereros defeated.

4

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Source: Resistance struggle 1904 by Klaus Dierks

Source data:

Book sources:

1. Let us die fighting (Drechsler, 1966)

2. The revolt of the Hereros (Bridgman, 1981)

3. South West Africa under German rule (Bley, 1971)

5

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

• References

Websites and online articles:

1. Chronology of the Namibian history (Dierks, 2000)

2. Herero Uprising 11 January 1904 (Namibia-10n1, 2013)

6

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Historical

Documents

Document Pre-Processing

Gazetteer Creation

Contextual Information

Extraction

Trajectory & Location

Event Extraction

Spatial & Temporal

Gazetteers

Text Processing

Language

Processing

Gazetteer

Matching

Annotation

Results

7

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Gazetteer Creation:

What do we want?

• Temporal expressions

• Spatial expressions

• Attributive information (Person’s names)

Spatial Gazetteer

• ANNIE gazetteer

• List of place names – 3859 place names

8

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Gazetteer Creation:

Temporal Gazetteer

• JAPE grammar rule

• Date Expressions – 7 Pattern rules

No. Entity Pattern

1 Date June 1904

2 Date June 13

3. Date June 13, 1904

4. Date 13 June

5. Date 13 June 1904

6. Date 11.06

7. Date 11.06.1904

9

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

10

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Contextual IE:

Entity Extraction Pipeline

Person

Date

location

11

Spatio-temporal relationships

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

GATE annotation framework:

12

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

GATE Annotation Results:

13

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Trajectory & location events extraction

• Combine to Location event(Persons’ name, Location, Date)

• Chronological order – as per text document

• Write to PostgreSQL Database

• Produce individual trajectories

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

• References

263 location events

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

1. Location visit events – Location events in time

2. Individual trajectories – Moving points in time

3. Battle events – Location events in time

Historical Spatio-temporal data

Theory of a moving point in time

Modelling historical events in ArcGIS

18

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

We are interested in:

Space and existence in time Where & When?

Change in position & time

Spatial relationships in time

Spatio-temporal Cluster Analysis – January location events

“Where”

19

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

• Answers

20

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Spatio-temporal Cluster Analysis – January location events

Why?

Space – time cube Analysis – Monthly location events

(x, y, time) representation

Answers:

“Where”?

“When”?

21

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Trajectory representations

22

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

• References

Time –Aware Map

23

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Story Map Journal

24

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

25

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

Positional uncertainties

• Location names that do not exist

• Approximated locations

• Uncertain geographic locations

Temporal uncertainties

• Uncertain duration of events

• Range of dates

Uncertainties in historical data

Conclusions

• Approach used successfully extracted spatial, temporal and attributive

• Provide basis for structured data – Interactive visual history teaching

systems

• Support Domain specific extractions

26

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations • Extracted data – small scale, poor temporal density – Limited support

• Trajectory points connecting distant locations at discrete times

• ArcGIS online provides good Cartographic visualization tools

Therefore, recommend:

• Development of time query functions.

• Development of trajectory representation functions.

• Development of functions to estimate time between moving

points.

• Use of Existing Geo& temporal taggers

27

• Introduction

• Case Study

• Source Data

• IE Workflow

• Results

• Conclusions &

Recommendations

THANK YOU FOR YOUR ATTENTION!

28