Unlocking Environmental Narratives

Front matter

Unlocking Environmental Narratives

Towards Understanding Human Environment Interactions through Computational Text Analysis

Ross S. Purves, Olga Koblet and Benjamin Adams

Chapter 2: Interdisciplinary Perspectives on Environmental Narratives

Ross S. Purves, Olga Koblet, Benjamin Adams, Katrin Anna Lund, Karen Jones, Sarah Luria, Flurina Wartmann, Graham Fairclough and Gabriel Viehhauser

Chapter 3: Analysing Environmental Narratives Computationally Ross S. Purves, Olga Koblet and Benjamin Adams

3.1 3.2 3.3 3.4

3.5 3.6

Narrative Forms

Building a Corpus

Copyright and Ethics

Corpus Linguistics and Natural Language Processing

3.4.1 Pre-processing and encoding natural language as features 3.4.2 Classification

Where to Next?

Suggested Readings vii iv CONTENTS

Part II: Case Studies

Chapter 4: Showing the Potential of Computational Analysis to Support Environmental Narrative Research Ross S. Purves, Olga Koblet and Benjamin Adams

4.1 Case Study Summaries

Chapter 5: Glacial Narratives: How Can They Be Captured? Katrin Anna Lund, Ludovic Moncla and Gabriel Viehhauser

5.1 Living Creatures 5.2 Experiments 5.2.1 Corpus 5.2.2 Word frequencies 5.2.3 Diachronic keyness 5.3 Discussion

Chapter 6: Greening a Post-Industrial City: Applying Keyword Extractor Methods to Monitor a Fast-Changing Environmental Narrative

Sarah Luria and Ricardo Campos

6.1 Theoretical Approach

6.2 Sources

6.3 Method

6.4 Interpretations and Results

6.5 Discussion of Implications, Limitations, and Links to Other Work

Chapter 7: Best Practice for Forensic Fishing: Combining Text Processing with an Environmental History View of Historic Travel Writing in Loch Lomond, Scotland

Karen Jones, Diana Maynard and Flurina Wartmann

7.1 Finding Relevant Sources and Compiling a Corpus of Digitised

Travel Descriptions

7.2 Forensic Fishing in Digital Text Sources Using Natural Language

Processing 7.2.1 Processing digitised historic texts in GATE to identify landscape terms and place names

7.2.2 Identifying relevant texts containing landscape descriptions of the Loch Lomond area

7.2.3 Application example 1: Tracing the emergence of touristic hotspots in Loch Lomond and the Trossachs

95 98 98 101 101 105

111 112 113 116 123

CONTENTS v

7.2.4 Application example 2: Words about landscape - Analysing historic landscape descriptions through term co-occurrence 7.2.5 Challenges in applying natural language processing to an analysis of historic travel writing 7.3 Reading the Trails as an Environmental Historian 7.4 Concluding Thoughts

Chapter 8: The Wild Process: Constructing Multi-Scalar Environmental Narratives Joanna E. Taylor and Benjamin Adams

8.1 Challenging the ‘Lone Enraptured Male’

8.2 Processing Literary Diversity on Rannoch Moor

8.3 Walking Rannoch Moor with Robert Macfarlane and Linda Cracknell

8.4 The Country Diary

8.5 Conclusion

Chapter 9: Inferring Value: A Multiscalar Analysis of Landscape Character Assessments

Joanna E. Taylor, Meladel Mistica, Graham Fairclough and Timothy Baldwin

9.1 Methodology 9.1.1 Creating a corpus 9.1.2 Parsing the corpus 9.1.3. Multiscalar approaches to LCAs 9.1.3.1 Topic modelling 9.1.3.2 Close reading 9.2 Results and Discussion 9.2.1 Assessing “Value” by integrating knowledge and technology 9.2.2 The value of character 9.2.3. Natural value and valuing nature 9.3 Conclusions and Further Work

Chapter 10: Interpreting Natural Spatial Language in a Fictional Text: Analysing Natural and Urban Landscapes in Mary Shelley’s Frankenstein

Tobias Zuerrer

10.1 Documenting the Perception of a Landscape Using Text 10.2 Methodology 10.3 Results

151 153 157

163 165

169 171 176

198 200 202 vi CONTENTS

10.4

10.5 10.6

Discussion

10.4.1 Natural landscape features in Frankenstein

10.4.2 Urban landscape features in Frankenstein

10.4.3 The natural and the urban - a landscape dichotomy? Further Studies

Conclusion

Chapter 11: Discovering Spatial Referencing Strategies in Environmental Narratives Simon Scheider, Ludovic Moncla and Gabriel Viehhauser

11.1 11.2 11.3

11.5 11.6

Mountaineering in Scotland Referencing Strategies in Environmental Narratives Geoparsing Frames of Reference

11.3.1 Perdido parsing rules

11.3.2 FoR rules

Results and Quality of Geoparsing 11.4.1 Inter-annotator agreement 11.4.2 Quality on mountaineering texts 11.4.3 Precision on adventure novel Discussion and Future Work Conclusion

Chapter 12: Surveying the Terrain and Looking Forward Ross S. Purves, Olga Koblet and Benjamin Adams

204 204 206 206 207 208

2.1

2.2 2.3 2.4 2.5

3.1

3.2

3.3

3.4

5.1

5.2 5.3

6.1

6.2

6.3

List of Figures

Environmental features of the Strandir region. 2.1b by Claus Sterneck, all other pictures by Katrin Lund.

Djupavik. Picture by Claus Sterneck.

Environmental features of Djupavik. Pictures by Katrin Lund.

Spatial frames in the fairy tale ‘Hansel and Gretel’

Distinctive words related to the witch’s house in the fairy tale ‘Hansel and Gretel.

Example of a webpage with its HTML structure.

Source: https://www.theyworkforyou.com/debates/?id=2019 -03-28b.545.8#2¢545.11

Example concordances in the Hansard corpus.

Sample dependency tree.

Adjectives ranked in BNC and the Hansard corpus.

Articles and tokens per year in the three corpora.

Word clouds of the most frequent words in the corpora.

Most distinctive words for texts appearing before and after January 2000.

Lincoln, William. History of Worcester, Massachusetts, from its earliest settlement to September, 1836 : with various notices relating to the history of Worcester County. Worcester, 1862, p. 291. Accessed through archive.org 14 Oct. 2019. Framed text shows material excerpted for corpus selection.

YAKE! word cloud (n = 3) Mary Fell, The Prophecy, from The Persistence of Memory (1984). Github filename: 7 - 1984_Mary Fell, The Prophecy (Poem).

YAKE! word cloud (n = 3) scroll of corpus.

19 21 22 37

49 60 62 62 100 102

123 125 viii

7.1

7.2

7.3

7.4

7.5

7.6

7.7

11.3

LIST OF FIGURES

Map of Loch Lomond and The Trossachs National Park study area (base map credit: Wikimedia user Nilfanion, map created using Ordnance Survey open data CC-BY-SA-3.0).

User interface of the GATE toolkit highlighting the annotated terms (landscape terms and Loch Lomond place names for settlements and water bodies — see Annex 1 & 2 for full list of search terms used).

Places within the study area mentioned in the “Guide to Romantic Scenery’ numbered according to first occurrence. (base map credit: Wikimedia user Nilfanion, map created using Ordnance Survey open data CC-BY-SA-3.0).

Places within the study area mentioned in the ‘Keddie and Gray numbered according to first occurrence. (base map credit: Wikimedia user Nilfanion, map created using Ordnance Survey open data CC-BY-SA-3.0).

Sequence of places names in a subsection of “Keddie and Gray’ scaled according to the number of mentions. Basemap © OpenStreetMap contributors.

Contemporary view of Loch Achray (photograph by F. Wartmann).

Frequencies of co-occurrences for beautiful, picturesque and sublime with landscape terms.

Concordance plots created in AntConc for T and ‘we in two of our texts.

Description of natural landscape features in Frankenstein. Description of urban landscape features in Frankenstein. Spatial word clouds of two landscape features.

Rannoch Moor on Open Street Map (OSM). © Open Street Map contributors.

Illustration of types of referencing strategies used in environmental narratives. Target locations are indicated in turquoise, ground objects in grey. See text for the meaning of labels.

Transducer for parsing Euclidean frames.

168 203 204 207

216 219

LIST OF FIGURES ix

11.4 Precision of detecting different types of reference frames in the first 3311 sentences of the 1886 novel Kidnapped, including ZF (‘zonal’), TF (‘topo’), LCF (‘linear’), EF (‘eucl’) and BF (‘betw’) frames. 225

11.5 Distribution of keywords and modifiers over true- and false-positive frame annotations. 226

1.1 3.1 3.2

3.3

3.4

3.5

6.1

6.2 6.3

6.4

7.1

7.2

7.3

8.1

8.2

List of Tables

Interpretation of the text Glencoe Road.

Basic counts for corpus.

20 most frequent terms in corpus before and after filtering for stop words.

Ten most frequent terms co-occurring with landscape ranked by frequency and mutual information.

Relation between Kappa statistics and strength of agreement as proposed by Landis and Koch (1977).

Confusion matrix of the first annotation of the Forestry Commission text into three classes: negative, neutral, positive.

Sarah’s Evaluation of YAKE! Results - sources linked, where available, online in footnotes.

‘Good’ Rating: Subject + Argument Words Extracted.

‘Sufficient’ Rating: Subject + Some Argument Words Extracted.

‘Insufficient’ Rating: Subject but No Argument Words Extracted.

Number of annotated terms identified through gazetteer lookup.

picturesque: Number of co-occurrences for landscape terms with the descriptor.

Example of digitised and original text in the Guide to the romantic scenery of Loch-Lomond, Loch-Ketturin, the Trosachs, James Lumsden, (1831).

Collocated words with ‘Rannoch by organised authors’ genders.

Counts of articles in the Country Diary corpus by category.

10 58

117 119

166 172 xii LIST OF TABLES

8.3 8.4 8.5 8.6

Naive Bayes results on a balanced set of 2770 articles. Naive Bayes results on a balanced set of 120 unique authors. Counts of personal pronouns per article.

Collocated words for the pronouns T and ‘we’ in the different corpora.

Criteria for assessing PDF Extractors.

Seed terms (note use of wild cards) retrieved in Mary Shelley’s Frankenstein.

Seed terms related to natural landscapes.

Seed terms related to urban landscapes.

Precision and recall of the first annotation round compared to the gold standard.

Precision and recall of the second annotation round compared to the gold standard.

Number of errors and evaluation scores for our gold standard corpus.

173 173 174

175 186

200 202 203

Notes about the book

About the cover

The cover image tells a story - it shows glacial striations - scratches left on bedrock by a glacier that has now receded. Such marks can be found in many locations across the world, but this site - the Rhone Glacier in Switzerland is special - since it has been visited by tourists, and written about, for more than 150 years. If we were to turn 180 degrees, then we would see the rapidly receding remains of the glacier where tourists, after paying an entry fee, can descend into ice grottos laboriously carved out of the ice each spring. We would also see thermal blankets, draped across the glacier in an attempt to slow its melt and preserve the grottos for a few more years. This narrative can be pieced together by reading newspaper stories, travel diaries and even fictional writing about this site, and analysing such material at scale, using computers, is the subject of this book.

Code related to this book

We have included the code and public domain data used in the introductory and case study chapters of this book in the following GitHub repository: https://github.com/Unlocking-Environmental-Narratives/Unlocking- Environmental-Narratives-Book, DOI: 10.5281/zenodo0.6766825.

Permission

Stanzas from the poem “The Prophecy” by Mary Fell have been used with permission of the author. All rights reserved for all elements of the poem.

Contributors

Department of Computer Science and Software Engineering, University of Canterbury, Christchurch, New Zealand

Email: benjamin.adams@canterbury.ac.nz

Web: https://www.canterbury.ac.nz/engineering/contact-us/people/ ben-adams.html

Benjamin Adams is an Associate Professor of Computer Science at the University of Canterbury in New Zealand. His expertise lies in the development and application of information retrieval and machine learning algorithms that aid the collection, management, analysis and use of large spatial data sets to better support decision-making in complex environments. A key driver of his research is the goal of developing computational methods that can translate unstructured and semi-structured data that are created initially for human communication, such as crowdsourced social media and other natural language text and images, into forms that aid geographic problem-solving.

Timothy Baldwin’”

‘Department of Natural Language Processing, The Mohamed bin Zayed University of Artificial Intelligence

School of Computing and Information Systems, The University of Melbourne, Australia

Timothy Baldwin is Associate Provost Academic and Student Affairs and Acting Head of the Department of Natural Language Processing, The Mohamed bin Zayed University of Artificial Intelligence in addition to being a Melbourne Laureate Professor in the School of Computing and Information Systems, The University of Melbourne. His primary research focus is on natural language processing, including social media analytics, deep learning and computational social science.

xvi CONTRIBUTORS

Ricardo Campos’?

*Ci2 - Smart Cities Research Center - ICT Departmental UNIT, Polytechnic Institute of Tomar, Portugal

7INESC TEC., Portugal

Email: ricardo.campos@ipt.pt

Web: http://www.ccc.ipt.pt/ ricardo/

Ricardo Campos is a Professor at the Polytechnic Institute of Tomar and a Senior Research at INESC TEC with more than 10 years of experience in Information Retrieval (IR) and Natural Language Processing (NLP). He is the leading author of the highly impactful Yake! keyword extractor toolkit. His current research focuses on developing methods concerned with the process of narrative extraction from texts. He is particularly interested in practical approaches regarding the relationship behind entities, events and temporal aspects, as a means to make sense of unstructured data.

Graham Fairclough

McCord Centre for Landscape, Newcastle University, UK

Email: graham.fairclough@newcastle.ac.uk

Web: https://www.ncl.ac.uk/hca/people/profile/grahamfairclough.html

Graham Fairclough is an archaeologist and historian with research interests in heritage and landscape studies. Until 2012 he worked in the UK government agency English Heritage, and is presently a research member of the McCord Centre for Landscape at Newcastle University (UK). He has co-edited the interdisciplinary journal Landscapes since 2011 and has published widely on heritage, landscape and sustainability. He coordinated the EU CHeriScape network (2012-2018) and is currently a member of the EU MSC HERILAND project (2018-2022) and the Anglo-German DFG/AHRC project Devastation & Relocation.

Karen Jones

School of History, University of Kent, UK

Email: k.rjones@kent.ac.uk

Web: https://www.kent.ac.uk/history/people/400/jones-karen ORCID: https://orcid.org/0000-0002-4787-6830

Twitter: @drkarenjones

Professor Karen Jones is an environmental historian at the University of Kent with a particular interest in place, storytelling and environmental change over time. She has written extensively on parks and conservation landscapes and on species entanglements between humans and other animals, especially wolves. Her current research focuses on city parks, health and urban metabolism, with a particular focus on recovering the voice of marginal communities across a global geography of urban greening.

CONTRIBUTORS xvii

Department of Geography, University of Zurich, Switzerland E-mail: olga.koblet@geo.uzh.ch

Olga Koblet is a former postdoctoral researcher at the Geocomputation Group of the University of Zurich. Her focus lies in the use of unstructured text to better understand landscape and landscape perception. In particular, she has worked on new approaches to extracting descriptions of perception from unstructured text and linking these to landscape character assessment.

Katrin Anna Lund

Institute of Life and Environmental Sciences, University of Iceland E-mail: kl@hi.is

Web: https://www.hi.is/starfsfolk/kl

Katrin Anna Lund is Professor of Geography and Tourism, Institute of Life and Environmental Sciences at the University of Iceland. She is a social anthropologist with a PhD from the University of Manchester. She has done research in Spain, Scotland and Iceland, as well as in Northern Norway and Finnish Lapland. Her research has focused on tourism, travel and the perception of landscape, but landscape studies have been central to her work on travel and tourism, with a special emphasis on walking and narratives. She has been studying destination making in tourism for the past 10 years.

Sarah Luria

Department of English

College of the Holy Cross, Worcester, Massachusetts E-mail: sluria@holycross.edu

Sarah Luria is a professor of American Literature and former Director of Environmental Studies at the College of the Holy Cross. She studies how literature and stories more broadly impact a sense of place and relationship to the environment. Her current book project, The Story of My Street, gives an in-depth tour of the environmental history of the street where she lives. She recently directed and edited a film on the Native American history and ongoing identity of her college campus and surrounding community.

xviii CONTRIBUTORS

Diana Maynard

Dept of Computer Science, University of Sheffield, UK

Email: d.maynard@sheffield.ac.uk

Web: https://www.sheffield.ac.uk/dcs/people/research-staff/diana-maynard ORCID: https://orcid.org/0000-0002-1773-7020

Twitter: @dianamaynard

Diana Maynard is a Senior Research Fellow at the University of Sheffield with more than 30 years of experience in Natural Language Processing (NLP). Since 2000 she has been one of the key developers of the GATE NLP toolkit, leading work on Sheffields open-source multilingual text analysis tools. Her main research interests are in practical, multidisciplinary approaches to text and social media analysis, in a wide range of fields including sustainability and the environment, politics, climate change, natural disasters, human rights, law and journalism.

Meladel Mistica

Melbourne Data Analytics Platform, University of Melbourne, Australia Email: misticam@unimelb.edu.au

Web: https://findanexpert.unimelb.edu.au/profile/3575-mel-mistica ORCID: https://orcid.org/0000-0001-6147-120X

Twitter: @undergoer

Meladel Mistica is a researcher at the Melbourne Data Analytics Platform (MDAP), University of Melbourne. She has extensive industry and academic experience in natural language processing (NLP). Her role at MDAP is focused on enabling and guiding academic colleagues from Humanities, Arts and Social Sciences in adopting computational and digital methods in their research.

Ludovic Moncla

Univ Lyon, INSA Lyon, CNRS, UCBL, LIRIS, UMR5205, F-69622, France E-mail: ludovic.moncla@insa-lyon.fr

Web: https://ludovicmoncla.github.io

Twitter: @MonclaLudovic

Ludovic Moncla is associate professor in Computer Science at INSA Lyon and at the LIRIS laboratory. His research focuses on multidisciplinary aspects of Natural Language Processing, Information Retrieval, Data Mining, Digital Humanities and Geographic Information Science. He develops methods for geoparsing and geocoding texts in order to solve digital and societal issues related to the territory and the exploitation of textual data.

CONTRIBUTORS xix

Ross Purves’

‘Department of Geography, University of Zurich, Switzerland "URPP Language and Space, University of Zurich, Switzerland E-mail: ross.purves@geo.uzh.ch

Web: https://www.geo.uzh.ch/ rsp

Twitter: @GCUZH

Ross Purves is a professor of Geographic Information Science at the Department of Geography, University of Zurich. His research aims to address societally relevant research questions, with the fundamental aim of making theoretical, thematic and methodological contributions to Geographic Information Science. He is particularly interested in developing methods and answering questions through the use of unstructured information, often in the form of text.

Simon Scheider

Department of Human Geography and Spatial Planning, Utrecht University, the Netherlands

E-mail: s.scheider@uu.nl

Simon Scheider is an assistant professor in Geographic Information Science at the Department of Human Geography and Spatial Planning, Utrecht University. His research lies at the interface between conceptual modeling, geographic data analysis and knowledge extraction. He is particularly interested in understanding the concepts underlying spatio-temporal data, including reference systems, fields, places, spatial objects, events, trajectories and their relationship to human activities, and in practical methods for modelling and handling them with GIS, Artificial Intelligence and semantic technology.

Joanna E. Taylor

Department of English, American Studies, and Creative Writing, University of Manchester, UK

E-mail: joanna.taylor@manchester.ac.uk

Web: https://www.research.manchester.ac.uk/portal/joanna.taylor.html ORCID: https://orcid.org/0000-0001-8597-0097

Twitter: @JoTayl0rd

Joanna E. Taylor is Presidential Fellow in Digital Humanities at the University of Manchester. Her research focuses on literary geographies of the long 19th century, spatial poetics and environmental humanities. Digital methodologies and technologies - particularly from Natural Language Processing and Geographical Information Science - extend the reach of this work into investigations into the relationships between a place’s cultures, ecologies and heritage.

xx CONTRIBUTORS

Gabriel Viehhauser

Department of Digital Humanities, University of Stuttgart, Germany E-mail: viehhauser@ilw.uni-stuttgart.de

Web: https://www.ilw.uni-stuttgart.de/institut/team/Viehhauser-00001/ ORCID: https://orcid.org/0000-0001-6372-0337

Gabriel Viehhauser is a professor for Digital Humanities at the University of Stuttgart. With a background in German medieval literary studies, his research focuses on digital editions and digital text analysis as well as on a combination of both fields. He is particularly interested in digital narratology and the modelling of space in narratives.

Flurina Wartmann

Department of Geography & Environment, University of Aberdeen, UK Email: flurina.wartmann@abdn.ac.uk

Web: https://www.abdn.ac.uk/geosciences/people/profiles/flurina.wartmann ORCID: https://orcid.org/0000-0003-4788-2963

Twitter: @DrFlurina

Flurina Wartmann is a Lecturer in Geography at the University of Aberdeen. Her research interests revolve around the relationship between society and nature, including nature conservation, landscape restoration and rewilding, landscape assessment, cultural landscape values and sense of place. In her research, she uses interdisciplinary approaches and a variety of data sources, including interviews, public surveys, social media photographs, digitised text sources and participatory Geographic Information Systems (GIS).

Tobias Ziirrer Email: tobias.zuerrer@ksa.sz.ch

Tobias Zürrer is a former English literature and geography student at the University of Zurich. He is interested in the conceptualisation of natural and urban landscapes in (historical) fictional texts.

CONTRIBUTORS xxi

Funding Acknowledgements

Benjamin Adams’ work on this book was partially supported by the Strategic Science Investment Fund from Government Funding, administered by the Ministry of Business Innovation and Employment, Aotearoa/New Zealand, through the grant ‘Beyond Prediction: explanatory and transparent data science (UOAX1932).

Ricardo Campos was financed by the ERDF European Regional Development Fund through the North Portugal Regional Operational Programme (NORTE 2020), under the PORTUGAL 2020 and by National Funds through the Portuguese funding agency, FCT - Fundação para a Ciência e a Tecnologia within project PTDC/CCI-COM/31857/2017 (NORTE-01-0145-FEDER-03185). This funding fits under the research line of the Text2Story project.

Sarah Lurias Research was partially funded by a Course Development Grant through the College of the Holy Cross, Worcester, Massachusetts.

Ross Purves gratefully acknowledges funding from the Swiss National Science Foundation (200020E_186389) and the University of Zurich.

Simon Scheider’s work was partially funded by the European Research Council (ERC) under the European Unions Horizon 2020 research and innovation programme (grant agreement No. 803498).

Foreword

This book started with an idea: we wanted to bring together a group of people, to work together on a theme that we considered exciting. All three of us are excited by the possibilities of analysing text as a tool for better understanding the world around us, and we were convinced that if we could bring the right people together, in the right place, we could do something worthwhile.

That idea turned into a workshop, with the contributors who have brought this book alive. In spring 2019, we travelled together to a beautiful mountain village in Stels, Switzerland, and spent four wonderful days brainstorming and discussing many of the ideas that came together to form this book. We left the workshop excited by what we had achieved, and with a concrete plan for a set of case studies to populate our original idea. We had a plan, and a timetable, and were confident that we could quickly bring this book to fruition. Of course, as academics, everything (almost) always takes longer than we planned. This book is no exception, and the events that we have all lived through since spring 2019 are part of the narrative of our lives. We have all come to appreciate our surroundings in ways we did not before, and many of us have not seen friends and family for a long time.

Throughout that time, this book has remained a dim flame, sometimes sputtering, and perhaps nearly going out. Occasionally, most often when we received a contribution, the flames got brighter, and we could once more see the end. Writing this material, and more importantly the reading that this work has involved, has once again convinced us of the power, the depth and the beauty of the written word. Of course, words alone do not suffice, and we are especially grateful to Laura Wysling for giving the book a consistent graphical look. All of the contributors have answered our many mails, dealt with our queries and helped us to get this project over the line, and we could not have done it without them. Thanks too are due to the reviewers of our book, and the editorial team at Ubiquity for constructive feedback along the way. And finally, thanks to Franziska Komossa for taking one last look at the text for us.

But the biggest thanks go to our families — over the last three years they have been the people with whom we have spent the most time, who've been there when we needed support, and who have reminded us of the other stuff in life that is important. Thanks.

Ross, Olga and Ben April, 2022

PART |

Themes, Resources and Methods

CHAPTER |