Case Based Reasoning Research and Development 26th International Conference ICCBR 2018 Stockholm Sweden July 9 12 2018 Proceedings Michael T. Cox

Download the full version of the textbook now at textbookfull.
com
Case Based Reasoning Research and Development

26th International Conference ICCBR 2018
Stockholm Sweden July 9 12 2018 Proceedings
Michael T. Cox
https://textbookfull.com/product/case-based-
reasoning-research-and-development-26th-
international-conference-iccbr-2018-stockholm-
sweden-july-9-12-2018-proceedings-michael-t-cox/
Explore and download more textbook at https://textbookfull.com

Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.
Case Based Reasoning Research and Development 28th

International Conference ICCBR 2020 Salamanca Spain June 8
12 2020 Proceedings Ian Watson
https://textbookfull.com/product/case-based-reasoning-research-and-
development-28th-international-conference-iccbr-2020-salamanca-spain-
june-8-12-2020-proceedings-ian-watson/
textbookfull.com
Case Based Reasoning Research and Development 22nd

International Conference ICCBR 2014 Cork Ireland September
29 2014 October 1 2014 Proceedings 1st Edition Luc
Lamontagne
https://textbookfull.com/product/case-based-reasoning-research-and-
development-22nd-international-conference-iccbr-2014-cork-ireland-
september-29-2014-october-1-2014-proceedings-1st-edition-luc-
lamontagne/
textbookfull.com
Artificial Intelligence for Knowledge Management 6th IFIP

WG 12 6 International Workshop AI4KM 2018 Held at IJCAI
2018 Stockholm Sweden July 15 2018 Revised and Extended
Selected Papers Eunika Mercier-Laurent
https://textbookfull.com/product/artificial-intelligence-for-
knowledge-management-6th-ifip-wg-12-6-international-workshop-
ai4km-2018-held-at-ijcai-2018-stockholm-sweden-july-15-2018-revised-
and-extended-selected-papers-eunika-mercier/
textbookfull.com
Oil Booms and Business Busts Why Resource Wealth Hurts

Entrepreneurs in the Developing World 1st Edition Nimah
Mazaheri
https://textbookfull.com/product/oil-booms-and-business-busts-why-
resource-wealth-hurts-entrepreneurs-in-the-developing-world-1st-
edition-nimah-mazaheri/
textbookfull.com
Teaching English in Context Beverly Derewianka
https://textbookfull.com/product/teaching-english-in-context-beverly-
derewianka/
textbookfull.com
ODI from BRIC Countries Firm level Evidence Valeria Gattai
https://textbookfull.com/product/odi-from-bric-countries-firm-level-
evidence-valeria-gattai/
textbookfull.com
Regional Conference on Science Technology and Social

Sciences RCSTSS 2016 Nor Azizah Yacob
https://textbookfull.com/product/regional-conference-on-science-
technology-and-social-sciences-rcstss-2016-nor-azizah-yacob/
textbookfull.com
Snowden Official Motion Picture Edition Kieran Fitzgerald
https://textbookfull.com/product/snowden-official-motion-picture-
edition-kieran-fitzgerald/
textbookfull.com
Short-Circuits in AC and DC Systems: ANSI, IEEE, and IEC

Standards J. C. Das
https://textbookfull.com/product/short-circuits-in-ac-and-dc-systems-
ansi-ieee-and-iec-standards-j-c-das/
textbookfull.com
Intermittent Fasting for Woman Over 50 1st Edition Sharon
Green
https://textbookfull.com/product/intermittent-fasting-for-woman-
over-50-1st-edition-sharon-green/
textbookfull.com
Michael T. Cox · Peter Funk
Shahina Begum (Eds.)
Case-Based Reasoning
LNAI 11156
Research
and Development
26th International Conference, ICCBR 2018
Stockholm, Sweden, July 9–12, 2018
Proceedings
123
Lecture Notes in Artificial Intelligence 11156
Subseries of Lecture Notes in Computer Science
LNAI Series Editors

Randy Goebel
University of Alberta, Edmonton, Canada
Yuzuru Tanaka
Hokkaido University, Sapporo, Japan
Wolfgang Wahlster
DFKI and Saarland University, Saarbrücken, Germany
LNAI Founding Series Editor

Joerg Siekmann
DFKI and Saarland University, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/1244
Michael T. Cox Peter Funk
•
Shahina Begum (Eds.)
Case-Based Reasoning
Research
and Development
26th International Conference, ICCBR 2018
Stockholm, Sweden, July 9–12, 2018
Proceedings
123
Editors
Michael T. Cox Shahina Begum
Wright State University Mälardalen University
Dayton, OH, USA Västeras, Sweden
Peter Funk
Mälardalen University
Västeras, Sweden
ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Artificial Intelligence
ISBN 978-3-030-01080-5 ISBN 978-3-030-01081-2 (eBook)
https://doi.org/10.1007/978-3-030-01081-2
Library of Congress Control Number: 2018955569
LNCS Sublibrary: SL7 – Artificial Intelligence
© Springer Nature Switzerland AG 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Inside this manuscript, you will discover the scientific papers from the 26th Interna-
tional Conference on Case-Based Reasoning (ICCBR 2018). The conference took place
during July 9–12, in Stockholm, Sweden. The ICCBR conference is the premier annual
meeting of the case-based reasoning (CBR) research community. The theme of ICCBR
2018 was “The Future of CBR” and was highlighted by several activities, including a
lively panel discussion of the same moniker.
ICCBR 2018 was held at the beautiful Stockholmsmässan Conference Center in
Stockholm, and it was co-located with the 27th International Joint Conference on
Artificial Intelligence (IJCAI) and the 23rd European Conference on Artificial Intel-
ligence (ECAI). Along with the other Federated AI Meeting (FAIM) conferences (i.e.,
AAMAS, ICML, and SoCS), immediately preceding IJCAI-ECAI-18, the ICCBR
conference started with a day-long social gathering (Monday), followed by a workshop
day and doctoral consortium (Tuesday), and concluded with two days (Wednesday and
Thursday) of major technical presentations and invited talks.
Tuesday included a lively panel discussion focused on the conference theme.
Moderated by Peter Funk (Mälardalen University, Sweden), the discussants (Kerstin
Bach - NTNU, Norway; David Leake - Indiana University, USA; Antonio Sánchez
Ruiz-Granados - Universidad Complutense de Madrid, Spain; Rosina Weber - Drexel,
USA; and Nirmalie Wiratunga - Robert Gordon University, UK) explored the future of
CBR and the vision it represents. In addition to the panel, the day included workshops,
a doctoral consortium for senior PhD students in the community, and ICCBR’s second
video competition. The themes of the six workshops were “Case-Based Reasoning and
the Explanation of Intelligent Systems,” “Synergies Between CBR and Machine
Learning,” “Computation and CBR,” “CBR and Deep Learning,” “Knowledge-Based
Systems in Computational Design and Media,” and “Reasoning About Time and
CBR.” We thank the chairs of these workshops for their persistence, the hard work that
produced the results, and their willingness to examine new topics related to the CBR
field of artificial intelligence.
Both Wednesday and Thursday began with a plenary address by noted researchers.
Professor Hector Munoz-Avila (Lehigh University, USA) described recent advances in
CBR and a new area of research called goal-driven autonomy. Professor Barry Smyth
(University College, Ireland) talked about how CBR is used in recommender systems
to promote positive choices that bear on physical fitness. These proceedings contain a
full-length invited paper for the first and an abstract of the second presentation.
During these two days, 39 papers were also presented as plenary oral or poster
presentations; they were selected among 77 submissions. Included in the proceedings,
these papers address many themes related to the theory and application of CBR and its
future direction. Topics included multiple papers on textual CBR and a number of
cognitive and human-oriented papers as well as hybrid research between CBR and
machine learning.
VI Preface
Taking a cue from last year’s invited talk on analogical reasoning by Henri Prade,
the main technical presentations on Wednesday included a Special Track on Compu-
tational Analogy. Although a first for ICCBR, the idea of a special track for the
conference is not new in the community; it represents an analogical adaptation of the
18 year string of CBR special tracks at FLAIRS (the yearly conference of the Florida
Artificial Intelligence Research Society). Analogy and CBR are closely related research
areas. Both employ prior cases to reason in complex situations with incomplete
information. Organized by Irina Rabkina (Northwestern University, USA),
Pierre-Alexandre Murena (Telecom ParisTech, France), and Fadi Badra (Université
Paris, France), four technical papers had oral presentations in the special track, and four
analogy posters received acceptance. In future years, we hope the special track concept
enables us to publish maturing work that first appears in a workshop format or
otherwise grow the community and its outreach.
Here we wish to acknowledge the essential support of the many people who were
crucial to organizing ICCBR 2018. First, Mobyen Uddin Ahmed (Mälardalen
University, Sweden) did a wonderful job arranging the many details as local chair.
Second, Dustin Dannenhauer (Naval Research Lab, USA) performed the difficult task
of publicity chair. Along with Mobyen Ahmed and Shamsul Alam (both of Mälarda-
len), Dustin managed the development and maintenance of the conference website and
social media connections. Mirjam Minor (Goethe University, Germany) managed all
six workshops, from the call for workshop proposals to their organization at the
conference and the separate proceedings. Cindy Marling (Ohio University, USA) and
Antonio A. Sánchez Ruiz Granados (Universidad Complutense de Madrid, Spain) were
in charge of the doctoral consortium, an event that is critical to the future of CBR
research and in line with this year’s conference theme. Nirmalie Wiratunga (Robert
Gordon) served as sponsorship chair, significantly increasing financial sponsorships.
Finally, Michael Floyd (Knexus Research, USA) and Brian Schack (Indiana Univer-
sity, USA) organized the Video Competition (the second of a new ICCBR tradition).
We are also grateful to the members of the Advisory Committee, who provided
invaluable guidance when called upon, and to the members of the ICCBR Program
Committee (and additional reviewers), who carefully reviewed the paper submissions.
We also give thanks to the local support group at Mälardalen University (Sweden) and
students from Wright State University (USA) for their volunteer work before, during,
and after the conference. Additionally, we would like to thank all the authors who
submitted the interesting and technically detailed papers you see herein and therefore
contributed to its success, as well as the conference attendees, who made these days
memorable for everyone.
Most importantly, we are very grateful for the support of our many sponsors,
including the National Science Foundation, Smart Information Flow Technologies, the
journal Artificial Intelligence, Brains4AI, British Telecon, Knexus Research, Springer,
and the Telenor-NTNU AI Laboratory. Given the extra expenses this year associated
with a large commercial conference center, these sponsors made the difference between
fiscal success and a financial deficit.
Finally, it was indeed a splendid experience to be in Stockholm. The last time IJCAI
was in Stockholm, CBR had a significant presence. IJCAI 1999 had two sessions on
CBR, an applications session where a CBR paper won distinguished paper award, and
Preface VII
a day-long workshop entitled “Automating the Construction of Case-Based Reason-

ers.” This year, ICCBR brought the CBR community to Stockholm as an equal partner
with IJCAI and all of the other FAIM conferences. Who know what the lies open to us
in the future? To be part of this future, be sure to attend ICCBR 2019 in Otzenhausen,
Germany, and keep an eye on http://www.iccbr.org.
July 2018 Michael T. Cox

Peter Funk
Shahina Begum
Organization
Program Committee
Mobyen Uddin Ahmed Mälardalen University, Sweden
Kerstin Bach Norwegian University of Science and Technology,
Norway
Fadi Badra Université Paris 13, France
Shahina Begum Mälardalen University, Sweden
Ralph Bergmann University of Trier, Germany
Tarek Richard Besold City, University of London, UK
Isabelle Bichindaritz State University of New York at Oswego, USA
Derek Bridge University College Cork, Ireland
Hernan Casakin Ariel University, Israel
Sutanu Chakraborti Indian Institute of Technology Madras, India
Alexandra Coman Capital One
Antoine Cornuéjols AgroParisTech, France
Michael Cox Wright State Research Institute, USA
Dustin Dannenhauer Naval Research Laboratory
Sarah Jane Delany Dublin Institute of Technology, Ireland
Jean-Louis Dessalles Telecom ParisTech, France
Belen Diaz-Agudo Universidad Complutense de Madrid, Spain
Odd Erik Norwegian University of Science and Technology,
Norway
Thomas Eskridge Florida Institute of Technology, USA
Mark Finlayson Florida International University, USA
Tesca Fitzgerald Georgia Institute of Technology, USA
Michael Floyd Knexus Research
Kenneth Forbus Northwestern University
Scott Friedman SIFT
Peter Funk Mälardalen University, Sweden
Ashok Goel Georgia Institute of Technology, USA
Pedro González Calero Universidad Politécnica de Madrid, Spain
Vahid Jalali Indiana University Bloomington, USA
Stelios Kapetanakis University of Brighton, UK
Mark Keane UCD Dublin, Ireland
Joseph Kendall-Morwick University of Central Missouri, USA
Kai-Uwe Kuehnberger University of Osnabrück, Institute of Cognitive
Science, Germany
Luc Lamontagne Laval University, Canada
Philippe Langlais University of Montreal, Canada
David Leake Indiana University Bloomington, USA
X Organization
Yves Lepage Waseda University, Japan

Jean Lieber LORIA, Inria Lorraine, France
Ramon Lopez De Mantaras IIIA - CSIC
Abhijit Mahabal Google
Cindy Marling Ohio University, USA
Stewart Massie Robert Gordon University, UK
Clifton McFate QRG
Mirjam Minor Goethe University Frankfurt, Germany
Stefania Montani University of Piemonte Orientale, Italy
Hector Munoz-Avila Lehigh University, USA
Pierre-Alexandre Murena Telecom ParisTech, France
Emmanuel Nauer LORIA, France
Santiago Ontañón Drexel University, USA
Miltos Petridis Middlesex University London, UK
Marc Pickett Google
Enric Plaza IIIA-CSIC
Luigi Portinale University of Piemonte Orientale, Italy
Irina Rabkina Northwestern University
Juan Recio-Garcia Universidad Complutense de Madrid, Spain
Gilles Richard IRIT Toulouse, France
Jonathan Rubin PARC, A Xerox Company
Ute Schmid University of Bamberg, Germany
Steven Schockaert Cardiff University, UK
Barry Smyth University College Dublin, Ireland
Frode Soermo Verdande Technology
Antonio A. Sánchez-Ruiz Universidad Politécnica de Madrid, Spain
Ian Watson University of Auckland, New Zealand
Rosina Weber Drexel University, USA
David Wilson UNC Charlotte, USA
Nirmalie Wiratunga Robert Gordon University, UK
Additional Reviewers
Ahmed, Mobyen Uddin Kübler, Eric

Eyorokon, Vahid Martin, Kyle
Fitzgerald, Tesca Mathew, Ditty
Funk, Peter Rugaber, Spencer
Grumbach, Lisa Wijekoon, Anjana
Herold, Miriam Zeyen, Christian
Klein, Patrick
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
Contents
Invited Paper
Adaptive Goal Driven Autonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Héctor Muñoz-Avila
Main Technical Papers
Answering with Cases: A CBR Approach to Deep Learning. . . . . . . . . . . . . 15

Kareem Amin, Stelios Kapetanakis, Klaus-Dieter Althoff,
Andreas Dengel, and Miltos Petridis
CEC-Model: A New Competence Model for CBR Systems Based

on the Belief Function Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Safa Ben Ayed, Zied Elouedi, and Eric Lefèvre
The Case for Case Based Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Isabelle Bichindaritz
Case Based Reasoning as a Model for Cognitive Artificial Intelligence . . . . . 62

Susan Craw and Agnar Aamodt
Explainable Distributed Case-Based Support Systems:

Patterns for Enhancement and Validation of Design Recommendations . . . . . 78
Viktor Eisenstadt, Christian Espinoza-Stapelfeld, Ada Mikyas,
and Klaus-Dieter Althoff
Tangent Recognition and Anomaly Pruning to TRAP Off-Topic Questions

in Conversational Case-Based Dialogues . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Vahid B. Eyorokon, Pratyusha Yalamanchili, and Michael T. Cox
Combining Case-Based Reasoning with Complex Event Processing

for Network Traffic Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Manuel Grob, Martin Kappes, and Inmaculada Medina-Bulo
AI-VT: An Example of CBR that Generates a Variety of Solutions

to the Same Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Julien Henriet and Françoise Greffier
A Textual Recommender System for Clinical Data . . . . . . . . . . . . . . . . . . . 140

Philipp Andreas Hummel, Frank Jäkel, Sascha Lange,
and Roland Mertelsmann
XII Contents
Harnessing Hundreds of Millions of Cases: Case-Based Prediction

at Industrial Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Vahid Jalali and David Leake
Case Base Elicitation for a Context-Aware Recommender System . . . . . . . . . 170

Jose Luis Jorro-Aragoneses, Guillermo Jimenez-Díaz,
Juan Antonio Recio-García, and Belén Díaz-Agudo
The SECCO Ontology for the Retrieval and Generation

of Security Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
Andreas Korger and Joachim Baumeister
Exploration vs. Exploitation in Case-Base Maintenance:

Leveraging Competence-Based Deletion with Ghost Cases . . . . . . . . . . . . . . 202
David Leake and Brian Schack
Dynamic Detection of Radical Profiles in Social Networks Using Image

Feature Descriptors and a Case-Based Reasoning Methodology. . . . . . . . . . . 219
Daniel López-Sánchez, Juan M. Corchado,
and Angélica González Arrieta
Segmentation of Kidneys Deformed by Nephroblastoma Using

Case-Based Reasoning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Florent Marie, Lisa Corbat, Thibault Delavelle, Yann Chaussy,
Julien Henriet, and Jean-Christophe Lapayre
FITsense: Employing Multi-modal Sensors in Smart Homes

to Predict Falls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Stewart Massie, Glenn Forbes, Susan Craw, Lucy Fraser,
and Graeme Hamilton
Embedded Word Representations for Rich Indexing: A Case Study

for Medical Records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
Katherine Metcalf and David Leake
Case-Based Data Masking for Software Test Management . . . . . . . . . . . . . . 281

Mirjam Minor, Alexander Herborn, and Dierk Jordan
A CBR Approach for Imitating Human Playing Style in Ms. Pac-Man

Video Game . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
Maximiliano Miranda, Antonio A. Sánchez-Ruiz, and Federico Peinado
Perks of Being Lazy: Boosting Retrieval Performance . . . . . . . . . . . . . . . . . 309

Mehmet Oğuz Mülâyim and Josep Lluís Arcos
Bayesian-Supported Retrieval in BNCreek: A Knowledge-Intensive

Case-Based Reasoning System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
Hoda Nikpour, Agnar Aamodt, and Kerstin Bach
Contents XIII
Personalised Human Activity Recognition Using Matching Networks . . . . . . 339

Sadiq Sani, Nirmalie Wiratunga, Stewart Massie, and Kay Cooper
Why Did Naethan Pick Android over Apple? Exploiting Trade-offs

in Learning User Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
Anbarasu Sekar, Devi Ganesan, and Sutanu Chakraborti
An Analysis of Case Representations for Marathon Race Prediction

and Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
Barry Smyth and Pádraig Cunningham
Dynamic Case Bases and the Asymmetrical Weighted

One-Mode Projection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
Rotem Stram, Pascal Reuss, and Klaus-Dieter Althoff
Novel Object Discovery Using Case-Based Reasoning and Convolutional

Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
J. T. Turner, Michael W. Floyd, Kalyan Moy Gupta, and David W. Aha
Modelling Similarity for Comparing Physical Activity

Profiles - A Data-Driven Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
Deepika Verma, Kerstin Bach, and Paul Jarle Mork
Investigating Textual Case-Based XAI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431

Rosina O. Weber, Adam J. Johs, Jianfei Li, and Kent Huang
Improving kNN for Human Activity Recognition with Privileged Learning

Using Translation Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448
Anjana Wijekoon, Nirmalie Wiratunga, Sadiq Sani, Stewart Massie,
and Kay Cooper
Considering Nutrients During the Generation of Recipes

by Process-Oriented Case-Based Reasoning . . . . . . . . . . . . . . . . . . . . . . . . 464
Christian Zeyen, Maximilian Hoffmann, Gilbert Müller,
and Ralph Bergmann
An Effective Method for Identifying Unknown Unknowns

with Noisy Oracle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
Bo Zheng, Xin Lin, Yanghua Xiao, Jing Yang, and Liang He
Special Track: Computational Analogy
On the Role of Similarity in Analogical Transfer . . . . . . . . . . . . . . . . . . . . 499

Fadi Badra, Karima Sedki, and Adrien Ugon
Predicting Preferences by Means of Analogical Proportions . . . . . . . . . . . . . 515

Myriam Bounhas, Marc Pirlot, and Henri Prade
XIV Contents
A Fast Mapper as a Foundation for Forthcoming Conceptual

Blending Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532
João Gonçalves, Pedro Martins, and Amílcar Cardoso
Production of Large Analogical Clusters from Smaller Example Seed

Clusters Using Word Embeddings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548
Yuzhong Hong and Yves Lepage
Case-Based Translation: First Steps from a Knowledge-Light Approach

Based on Analogy to a Knowledge-Intensive One . . . . . . . . . . . . . . . . . . . . 563
Yves Lepage and Jean Lieber
Making the Best of Cases by Approximation, Interpolation

and Extrapolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 580
Jean Lieber, Emmanuel Nauer, Henri Prade, and Gilles Richard
Opening the Parallelogram: Considerations on Non-Euclidean Analogies . . . . 597

Pierre-Alexandre Murena, Antoine Cornuéjols,
and Jean-Louis Dessalles
Experiments in Learning to Solve Formal Analogical Equations . . . . . . . . . . 612

Rafik Rhouma and Philippe Langlais
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627

Invited Paper
Adaptive Goal Driven Autonomy
Héctor Muñoz-Avila(B)
Computer Science and Engineering, Lehigh University, 19 Memorial Drive West,

Bethlehem, PA 18015-3084, USA
[email protected]
Abstract. Goal-driven autonomy (GDA) is a reflective model of goal

reasoning combining deliberative planning and plan execution monitor-
ing. GDA’s is the focus of increasing interest due in part to the need to
ensure that autonomous agents behave as intended. However, to perform
well, comprehensive GDA agents require substantial domain knowledge.
In this paper I focus on our work to automatically learn knowledge used
by GDA agents. I also discuss future research directions.
1 Introduction
Goal-driven autonomy (GDA) is a reflective model of goal reasoning combining
deliberative planning and plan execution monitoring. The key aspect of GDA
is the intermix between the agent’s observations (e.g., from sensors), goal rea-
soning, including formulating new goals, and acting according to the goals it is
pursuing. GDA is related to fields such as the actor’s view of planning [15] (i.e.,
a framework for interleaving planning and execution), online planning [20] (i.e.,
refining a planning solution while executing it in the environment), cognitive
systems [21] (e.g., agents endowed with self-reflection capabilities) and general
agency [4]. The key distinctive feature of goal reasoning agents is their capability
to adjust their goals, including changing their goals altogether while acting in an
environment. This in contrast to approaches such as replanning [14,36] where
the plan generated is modified due to changes in the environment while still
aiming to achieve the same goals. For example, in [26], we use guiding principles
we called motivators to select which goals to achieve. As indicated in [2], there
is an increasing interest in goal reasoning, in part, because of applications such
as UUVs’ control [29], air combat [13] and disaster response [33].
To perform well, comprehensive GDA agents require substantial domain
knowledge to determine expected states [24], identify and explain discrepancies
[22], formulate new goals [39], and manage pending goals [37]. This requires, for
example, the agent’s programmers to anticipate what discrepancies can occur,
identify what goals can be formulated, and define their relative priority.
In this paper, I will focus on our work to automatically learn knowledge used
by GDA agents performed over the past years. This includes:
This work is supported in part under ONR N00014-18-1-2009 and NSF 1217888.
c Springer Nature Switzerland AG 2018
M. T. Cox et al. (Eds.): ICCBR 2018, LNAI 11156, pp. 3–12, 2018.
https://doi.org/10.1007/978-3-030-01081-2_1
4 H. Muñoz-Avila
– Using reinforcement Learning to acquire planning, explanation and goal for-

mulation knowledge (Sect. 3).
– Using case-based reasoning techniques for learning expectations, and goals
(Sect. 4).
– Formulating a taxonomy of automatically generated expectations using
learned hierarchical task network representations (Sect. 5).
In Sect. 6, I will discuss ongoing research directions including computing

expectations beyond a linear sequence of actions.
2 The Goal Driven Autonomy Model
GDA agents generate a plan π achieving some goals g using a planning domain
Σ [5,24,25] (see Fig. 1). They continually monitor the environment to check if
the agent’s own expectations x are met in the state s. When a discrepancy D
is found between the agent’s expectations X and the environment s, the agent
Fig. 1. The goal-driven autonomy model [24, 25]

Adaptive Goal Driven Autonomy 5
generate a plausible explanation e for d; as a result of this explanation, a new

goal ĝ is generated, thereby restarting the cycle with g = ĝ.
If A denotes all actions in the domain and S denotes all states, the following
are the knowledge artifacts needed by GDA agents:
– Σ, which contains a function γ : S × A → S, indicating for each action a ∈ A,
and every state s ∈ S, what is the resulting state γ(s) when applying a in s.
– Π, a planning system. In addition to Σ, it may include additional knowledge
to generate plans such as HTN planning knowledge.
– G, the collection of all goals that can be pursued by the agents.
– Given an action a in a plan π, what are the expectations of when executing
a. This is dependent on where a occurs in π as well as the initial state from
which π started its execution.
– The explanation knowledge. For example, explanation knowledge may consist
of a collection of rules of the form D → e, indicating for a discrepancy D what
is its explanation e.
– Goal formulation knowledge. For example, goal formulation knowledge may
consist of a collection of rules of the form e → g̃, indicating for each explana-
tion, what is the next goal to achieve.
– Goal selection knowledge. In its most simple form it always selects to last
goal generated by Goal Formulator (i.e., g = g̃).
3 Learning Plan, Explanation and Goal Formulation

Knowledge with Reinforcement Learning
We explored the use of reinforcement learning (RL) techniques [35] to learn
planning knowledge for GDA. The assumptions are that the agent knows the
plausible actions As ⊂ A that can be taken in an state s ∈ S. But the problem
is that the agent doesn’t know, among those actions in As , which is the best or
even a good action to take. Typical of RL, the agent aims at learning the value
of each action a ∈ As in such a way that it maximizes its rewards. The result
of RL’s learning process is a policy π : S → A, indicating for every state, which
action π(s) to take [17].
We also took advantage of RL’s capabilities to both select plausible explana-
tions for a discrepancy [12] and to select possible goals for a given explanation
[18]. In both cases we formulate the selection problem as a Markov Decision Prob-
lem (see Table 1). The aim is to choose the best explanation among the possible
discrepancies. Thus we view each discrepancy as an state and each explanation
as an action. All other aspects of the GDA process being equal, if the overall
performance of the GDA agent improves, we will assign a positive reward to
the explanation selected and a negative reward otherwise. This means that the
agent could consider multiple plausible explanations for the same discrepancy
and over time it learns the value of each explanation (i.e., higher value means
more plausible explanation).
We did a similar approach to learn goal selection knowledge but this time
modeling the possible explanations as the states and the choice of a goal as the
6 H. Muñoz-Avila
Table 1. Modeling explanation selection and goal formulation as an MDP
MDP Actions States

Explanation generation All possible explanations All possible discrepancies
Goal formulation All possible goals All possible explanations
action. These were separate works and we never tried to simultaneously learn
explanation selection and goal formulation knowledge. We will retake this point
in the future work section.
We tested our approach in adversarial games. Specifically real-time strategy
(RTS) games. In these games players control armies to defeat an opponent.
These games add elements that make game playing challenging: armies consists
of different types of units. The outcome of unit versus unit combat follows a
paper-rock-scissors design: archers can easily defeat gryphons but are themselves
easily defeated by knights. In turn knights can be easily defeated by gryphons.
Thus, the best players are those who maintain a well balanced army and know
how to manage it during battle. Another aspect that makes RTS games even
more challenging is the fact that players make their moves synchronously thereby
rewarding players that can make decisions quickly.
In our experiments [12,17,18], the GDA agent is controlling a player that is
competing against various opponents, one at the time in each game instance.
These opponents used a variety of hard coded strategies. For testing we did the
typical leave-one-out strategy: given N opponents, the GDA agent trained by
playing against N − 1 opponents and was tested against the remaining opponent
not used in training. The results were compelling showing a significant improve-
ment by the learning GDA agent. The reward function was the difference in
score between the GDA player and its opponent. Thus, a positive score means
the GDA is defeating the opponent whereas a negative reward indicates that the
opponent is winning.
4 Learning Expectations and Goals with Case-Based

Reasoning
Case-based reasoning (CBR) is a problem solving method in which cases,

instances of previous problem-solving episodes, are reused to solve new prob-
lems [1]. One of the situations when it is desirable to use CBR is when there
are gaps in the knowledge preventing the formulation of a complete theory, one
that can be reasoned by using first-principles. This is precisely the situation we
encountered when computing the expectations after taking an action a in state s,
the agent does not have any knowledge about the expected state. Furthermore,
since the gaming domains are nondeterministic, meaning that after applying
an action a to an state s there could be many possible resulting states as and
the agent initially does not know which are those states and the probability that
anyone of them will be reached after applying action a in s. To solve this problem
we created a case base, CBX of expectations [18]:
CBX : S × A → 2S×[0,1]
This case base maps for each state-action pair (s, a) a probability distribution
of pairs (q, p) indicating the probability p of reaching state q. In a typical CBR
manner, each time the controller selects an action a from state s (i.e., as indicated
by a policy π as described in the previous section), then one of the following two
steps is performed:
– If q was never reached before when action a was selected from state s, then
(q, p) is stored in the CBX with p = 1/N(s,a) , where N(s,a) is the number of
times a has been selected when reaching state s.
– If q has been selected N(s,a,q) times then the value of p is updated to
p = (N(s,a,q) + 1)/N(s,a) .
CBR is also used to learn to associate a goal state sπ to a policy π

[18,19]. Following a policy from an starting state s0 produces an execution trace
s0 π(s0 ) · · · π(sn ) sn+1 . If an state q appears with at least some predefined fre-
quency f on the last k steps in the trace then q is assigned to be the goal of the
policy (f and k are parameters set by the user). Otherwise sn+1 is assigned to be
the goal. Since different chains may be generated for the same initial state and pol-
icy, we keep a probability distribution of goals akin to the way we kept a provability
distribution of expectations described before. Although we only annotate the pol-
icy with the goal having the highest probability.
The goals are used as described in the previous section in the goal formulation
procedure. When a goal gπ is selected, its associated policy π is executed. Since
our agents perform on-line learning, π changes over time, which may result in
gπ itself changing.
5 A Taxonomy of Expectations with HTN Planning
For this research thrust, we leverage on our previous work for automatically
learning hierarchical task networks (HTN) [16,40]. The learning problem can
be succinctly described as follows: given a collection of plans generated by some
(possibly unknown) agent and a collection of tasks T , learn HTN planning knowl-
edge in a manner that is consistent with the given collection of plans. The learned
HTN indicate how to decompose a task t ∈ T into subtasks t1 · · · tm in T and
the applicability conditions (i.e., the preconditions) under which this task decom-
position, t → t1 · · · tm , is valid. The learned HTNs can be used by the HTN
planner SHOP [27] to generate plans. That is, we set Π = SHOP and π is an
HTN plan.
Under the premise of using HTN planning for plan generation, we exam-
ine the notion of expectations for GDA agents. Expectations play an important
role in the performance of GDA agents [8]: if the expectations are too general,
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
8 H. Muñoz-Avila
Fig. 2. Generic HTN; actions are divided between those already executed and those
to execute; analogously, tasks are divided between those accomplished and those that
remain open
many unnecessary discrepancies will be found triggering unnecessary and pos-

sible detrimental GDA process. An example of a detrimental situation is the
following: suppose in an RTS game that the GDA is very close to destroying an
opponent’s town hall, thereby defeating the opponent. Suppose that at the same
time the player unexpectedly detects a large opponent’s force poised to take
over the player’s resource gathering operation (resources are needed to increase
a player’s army). This discrepancy may trigger a new goal to defend the resource
gathering operation, which in turn may deviate part of the army from attacking
the town hall and miss the chance of winning the game. If the expectations are
too narrow, discrepancies can be missed in situations were formulating a goal
would have been beneficial.
We studied the problem of eliciting expectations from plans [9] and from
HTN plans [8]. The basic premise is to divide the plan into two portions: the
actions already executed and the actions that remain to execute Fig. 2. SHOP
propagates forward the state and the next action to execute is checked against
the current state. Correspondingly, the task hierarchy can be divided between
those tasks that have been achieved because their underlying portion of the plan
is executed and those that are open (i.e., remain to be achieved) because some
of its underlying actions have not been achieved.
We explored 5 forms of expectations used in by GDA agents and more gen-
erally in goal reasoning agents:
– None. It checks the preconditions of the next action to execute. That is, if the
preconditions are applicable in the current state. This is useful, for example,
in the early stages of an RTS game when the player is focusing on building
its own economy. It typically focus on a predefined plan (this is called the
build order [28]). As soon as the preconditions of the next action become
applicable, it executes it. For example, building a peasant requires a certain
amount of gold; the player needs to wait until that amount is reached before
actually proceeding to build the peasant. All the following expectations check
the preconditions of the next action plus some additional conditions.
– Immediate. It checks if the effects of the last executed action has been
completed. This can be useful in highly reactive environments such as first-
person shooter games were for example the agent is checking if after jumping
it reaches the targeted platform.
– State. This is the most common form of expectations used in GDA. The
agent checks if the conditions in the state obtained by projecting from the
starting state after each action executed so far match the observations in
the environment. This is a natural way for GDA agent using a variety of
conditions in the state including symbolic [25], ontological [6] and numerical
[38]. It ensures that the part of the plan that remains to be executed, it is
still executable from the current state.
– Informed. This is a form we created [9]. Instead of projecting the whole
state, it projects only the accumulated effects from the actions executed so
far. The basic idea is that only the accumulated effects are the necessary
conditions that should be checked against the actual state. There might be
observations in the actual state that do not match the projected state, but if
those observations are not propagated conditions, there is no need to trigger
the GDA process. For example, if in an RTS game, the plan π calls for taking
a resource area in the center of the map, and while executing this plan it
detects some neutral units in a nearby location that pose no threat to the
plan, they can simply ignore them, without triggering an unnecessary GDA
cycle. In contrast, state expectations will trigger an iteration of the GDA
cycle every time any observation doesn’t match the expected state.
– informed-k. Informed-k expectations are informed expectations that are
checked every k steps, where k is an input parameter to the agent. These
expectations were created for situations when there are costs associated with
sensing if a condition in the expectations matches an actual observation in
the state [7]. For example, a vehicle might need to stop to turn the sensors, in
which case the cost can be measured in terms of the time it needs to remain
immobile.
Various experiments were made using variations of benchmark domains from

the literature such as the Mudworld [23] and the Arsonist [30] domains. In these
experiments all other components of the GDA agent were the same and the only
change was the type of expectations they were computing [7–9]. The following
observations were made: none and immediate expectations had no or very low
sensing costs but GDA agents frequently failed to achieve the goals it expected to
have achieved. GDA agents using state expectations achieved all goals that the
agent expected to achieve but had the highest sensing costs compared to agents
using other forms of expectations. Using informed expectations also achieve all
goals but at a much lower costs than when using state expectations. Depending
10 H. Muñoz-Avila
on the parameter k, Informed-k expectations also achieved all goals while having
the lowest cost.
6 Current and Future Research Directions
We are currently exploring a variety of directions. First, existing work on expec-

tations (ours and others) assume the plan π as a sequence of actions (possibly
augmented with a hierarchy of tasks). When the solution π is a policy, existing
research is looking at the execution trace from following the policy [17]; that is, a
sequence of actions. We are currently exploring goal regression for policies, which
unlike for the case of sequences of actions, cannot guarantee the minimal set of
necessary conditions as opposed to the deterministic case [32]. We are exploring
combining informed and goal regression expectations for the non-deterministic
case [31].
Second, we also will like to explore the definitions of different forms of expec-
tations for domains that include a combination of numeric and symbolic variables
and outcomes are dictated by (a possible unknown) probability distribution. This
includes situations when, for example, the agent’s gasoline consumption is deter-
mined using a probability distribution while at the same time it is ascertaining
if a message was sent or not.
Third, we will like to explore the interaction between the different compo-
nents of the GDA process. For example, how the learned planning knowledge
affects the expectations; in turn how discrepancies generated as a result of expec-
tations elicited in this particular way affect the kinds of explanations generated
and how this in turn determine the goals formulated. For this purpose we need
a common plan representation formalism. We recently adopted the hierarchi-
cal goal networks (HGNs) formalism [34] instead of hierarchical task networks
(HTN) for the planning formalism used by the planner Π. Unlike HTNs, HGNs
represent goals, not tasks, at all echelons of the hierarchy. This has been shown
to be useful to learn the hierarchies [3]; it also been shown to be particularly
suitable for execution monitoring [10,11] as the agent can determine directly if
a goal is achieved; unlike tasks in HTNs that do not have explicit semantics.
In HTNs, the only way to fulfill tasks is by executing the underlying plan gen-
erated from the HTN planning process. HGNs have the same expressiveness as
total-oder HTN planning [34].
Acknowledgements. This research is supported by ONR under grant N00014-18-1-

2009 and by NSF under grant 1217888. No work of this scope can be done by a single
person; I will like to thank the following external collaborators: David W. Aha and
David Wilson (Naval Research Laboratory), Michael T. Cox and Matthew Molineaux
(Wright State University); I will also like to thank the following (current and former)
students: Dustin Dannenhauer, Chad Hogg, Sriram Gopalakrishnan, Morgan Fine-
Morris, Noah Reifsnyder and Ulit Jaidee.
References
1. Aamodt, A., Plaza, E.: Case-based reasoning: foundational issues, methodological
variations, and system approaches. AI Commun. 7(1), 39–59 (1994)
2. Aha, D.W.: Goal reasoning: foundations emerging applications and prospects. AI
Mag. Under review (2018)
3. Choi, D., Langley, P.: Learning teleoreactive logic programs from problem solving.
In: Kramer, S., Pfahringer, B. (eds.) ILP 2005. LNCS (LNAI), vol. 3625, pp. 51–68.
Springer, Heidelberg (2005). https://doi.org/10.1007/11536314 4
4. Coddington, A.M., Luck, M.: A motivation-based planning and execution frame-
work. Int. J. Artif. Intell. Tools 13(01), 5–25 (2004)
5. Cox, M.T.: Perpetual self-aware cognitive agents. AI Mag. 28(1), 32 (2007)
6. Dannenhauer, D., Munoz-Avila, H.: LUIGi: a goal-driven autonomy agent reason-
ing with ontologies. In: Advances in Cognitive Systems (ACS 2013) (2013)
7. Dannenhauer, D.: Self monitoring goal driven autonomy agents. Ph.D. thesis,
Lehigh University (2017)
8. Dannenhauer, D., Munoz-Avila, H.: Raising expectations in GDA agents acting in
dynamic environments. In: IJCAI, pp. 2241–2247 (2015)
9. Dannenhauer, D., Munoz-Avila, H., Cox, M.T.: Informed expectations to guide
GDA agents in partially observable environments. In: IJCAI, pp. 2493–2499 (2016)
10. Dvorak, D.D., Ingham, M.D., Morris, J.R., Gersh, J.: Goal-based operations: an
overview. JACIC 6(3), 123–141 (2009)
11. Dvorak, D.L., Amador, A.V., Starbird, T.W.: Comparison of goal-based operations
and command sequencing. In: Proceedings of the 10th International Conference on
Space Operations (2008)
12. Finestrali, G., Muñoz-Avila, H.: Case-based learning of applicability conditions for
stochastic explanations. In: Delany, S.J., Ontañón, S. (eds.) ICCBR 2013. LNCS
(LNAI), vol. 7969, pp. 89–103. Springer, Heidelberg (2013). https://doi.org/10.
1007/978-3-642-39056-2 7
13. Floyd, M.W., Karneeb, J., Aha, D.W.: Case-based team recognition using learned
opponent models. In: Aha, D.W., Lieber, J. (eds.) ICCBR 2017. LNCS (LNAI),
vol. 10339, pp. 123–138. Springer, Cham (2017). https://doi.org/10.1007/978-3-
319-61030-6 9
14. Fox, M., Gerevini, A., Long, D., Serina, I.: Plan stability: replanning versus plan
repair. In: ICAPS, vol. 6, pp. 212–221 (2006)
15. Ghallab, M., Nau, D., Traverso, P.: The actor’s view of automated planning and
acting: a position paper. Artif. Intell. 208, 1–17 (2014)
16. Hogg, C., Muñoz-Avila, H., Kuter, U.: HTN-MAKER: learning HTNs with minimal
additional knowledge engineering required. In: Conference on Artificial Intelligence
(AAAI), pp. 950–956. AAAI Press (2008)
17. Jaidee, U., Muñoz-Avila, H., Aha, D.W.: Learning and reusing goal-specific policies
for goal-driven autonomy. In: Agudo, B.D., Watson, I. (eds.) ICCBR 2012. LNCS
1007/978-3-642-32986-9 15
18. Jaidee, U., Muñoz-Avila, H., Aha, D.W.: Integrated learning for goal-driven auton-
omy. In: Proceedings of the Twenty-Second International Joint Conference on Arti-
ficial Intelligence-Volume Volume Three, pp. 2450–2455. AAAI Press (2011)
19. Jaidee, U., Muñoz-Avila, H., Aha, D.W.: Case-based goal-driven coordination of
multiple learning agents. In: Delany, S.J., Ontañón, S. (eds.) ICCBR 2013. LNCS
1007/978-3-642-39056-2 12
12 H. Muñoz-Avila
20. Keller, T., Eyerich, P.: PROST: probabilistic planning based on UCT. In: ICAPS
(2012)
21. Langley, P.: The cognitive systems paradigm. Adv. Cogn. Syst. 1, 3–13 (2012)
22. Molineaux, M.: Understanding What May Have Happened in Dynamic, Partially
Observable Environments. Ph.D. thesis, George Mason University (2017)
23. Molineaux, M., Aha, D.W.: Learning unknown event models. In: AAAI, pp. 395–
401 (2014)
24. Molineaux, M., Klenk, M., Aha, D.W.: Goal-driven autonomy in a navy strategy
simulation. In: AAAI (2010)
25. Muñoz-Avila, H., Jaidee, U., Aha, D.W., Carter, E.: Goal-driven autonomy with
case-based reasoning. In: Bichindaritz, I., Montani, S. (eds.) ICCBR 2010. LNCS
1007/978-3-642-14274-1 18
26. Muñoz-Avila, H., Wilson, M.A., Aha, D.W.: Guiding the ass with goal motivation
weights. In: Goal Reasoning: Papers from the ACS Workshop, pp. 133–145 (2015)
27. Nau, D.S., Cao, Y., Lotem, A., Muñoz-Avila, H.: SHOP: simple hierarchical
ordered planner. In: Dean, T. (ed.) International Joint Conference on Artificial
Intelligence (IJCAI), pp. 968–973. Morgan Kaufmann, August 1999
28. Ontanón, S., Synnaeve, G., Uriarte, A., Richoux, F., Churchill, D., Preuss, M.:
A survey of real-time strategy game ai research and competition in starcraft. IEEE
Trans. Comput. Intell. AI Games 5(4), 293–311 (2013)
29. Oxenham, M., Green, R.: From direct tasking to goal-driven autonomy for
autonomous underwater vehicles. In: 5th Goal Reasoning Workshop at IJCAI 2017
(2017)
30. Paisner, M., Maynord, M., Cox, M.T., Perlis, D.: Goal-driven autonomy in dynamic
environments. In: Goal Reasoning: Papers from the ACS Workshop, p. 79 (2013)
31. Reifsnyder, N., Munoz-Avila, H.: Goal reasoning with goldilocks and regression
expectations in nondeterministic domains. In: 6th Goal Reasoning Workshop at
IJCAI/FAIM 2018 (2018)
32. Reiter, R.: The frame problem in the situation calculus: a simple solution (some-
times) and a completeness result for goal regression. In: Lifschitz, V. (ed.) Artificial
Intelligence and Mathematical Theory of Computation: Papers in Honor of John
McCarthy, (Ed.). Academic Press (1991)
33. Roberts, M., et al.: Goal reasoning to coordinate robotic teams for disaster relief.
In: Proceedings of ICAPS-15 PlanRob Workshop, pp. 127–138. Citeseer (2015)
34. Shivashankar, V.: Hierarchical Goal Network Planning: Formalisms and Algorithms
for Planning and Acting. Ph.D. thesis, Department of Computer Science, Univer-
sity of Maryland (2015)
35. Sutton, R.S., Barto, A.G.: Reinforcement Learning: An Introduction. MIT Press,
Cambridge (1998)
36. Warfield, I., Hogg, C., Lee-Urban, S., Munoz-Avila, H.: Adaptation of hierarchical
task network plans. In: FLAIRS Conference, pp. 429–434 (2007)
37. Weber, B.G., Mateas, M., Jhala, A.: Applying goal-driven autonomy to starcraft.
In: AIIDE (2010)
38. Wilson, M.A., McMahon, J., Aha, D.W.: Bounded expectations for discrepancy
detection in goal-driven autonomy. In: AI and Robotics: Papers from the AAAI
Workshop (2014)
39. Wilson, M.A., Molineaux, M., Aha, D.W.: Domain-independent heuristics for goal
formulation. In: FLAIRS Conference (2013)
40. Zhuo, H.H., Muñoz-Avila, H., Yang, Q.: Learning hierarchical task network
domains from partially observed plan traces. Artif. Intell. 212, 134–157 (2014)
Random documents with unrelated
content Scribd suggests to you:
courage, and concealing her poverty behind miraculous neatness and
cleanliness, never complaining of God, or of any mortal, refusing to talk of the
past, sometimes caressing children by stealth and leaving them as soon as one
looked at her, fearing, doubtless, that she might betray some emotion.
"I had not seen her for three years, when I came to ask you to paint my
portrait, and I was trying to find her address, which I was on the point of asking
you for when you mentioned her to me. Having arrived only the night before, I
did not know that she had at last attained success, celebrity, and a comfortable
income.
"It was only on finding her under such conditions, that it occurred to me that
that heart, so long crushed, might live again, and suffer—or be happy. Try to
make her happy, my dear Laurent, she has well earned it! And, if you are not
sure that you will not make her suffer, blow out your brains to-night rather than
return to her. That is all that I had to say to you."
"Stay," said Laurent, deeply moved; "this Comte de ——, is he still alive?"
"Unfortunately, yes. These men who drive other people to despair are always
in good health and escape all dangers. They never hand in their resignations;
why, this fellow recently had the presumption to send me a letter for Thérèse,
which I handed to her in your presence, which she treated as it deserved."
Laurent had thought of marrying Thérèse as he listened to Monsieur
Palmer's narrative. That narrative had produced a revolution in him. The
monotonous tone, the pronounced accent, and a few curious grammatical
lapses which we have not thought it worth our while to reproduce, had imparted
to it, in his auditor's vivid imagination, an indefinable something as strange and
terrible as Thérèse's destiny. That girl without kindred, that mother without
children, that wife without a husband, was surely doomed to an exceptionally
cruel fate! What depressing notions of love and life she must have retained! The
sphinx reappeared before Laurent's dazzled eyes. Thérèse unveiled seemed to
him more mysterious than ever; had she ever been consoled? could she be for
a single instant?
He embraced Palmer effusively, swore to him that he loved Thérèse, and
that, if he ever succeeded in winning her love, he would remember every hour
of his life the hour that had just passed and the story he had just heard. Then,
having promised not to betray his knowledge of Mademoiselle Jacques's history,
he went home and wrote:
"THÉRÈSE:
"Do not believe a word of all I have been saying to you these last two
months. Do not believe either what I said to you when you were afraid that I
would fall in love with you. I am not amorous, it is not that; I love you madly.
It is absurd, it is insane, it is wretched; but I, who thought that I never
should or could say or write to a woman the words I love you! find them too
cold and constrained to-day to express my feeling for you. I can live no
longer with this secret which is choking me, and which you will not guess. I
have tried a hundred times to leave you, to go to the end of the world, to
forget you. In an hour I am at your door, and very often, at night, consumed
with jealousy and almost frantic with rage against myself, I pray God to
deliver me from my torment by summoning this unknown lover in whom I do
not believe, and whom you invented to disgust me with the thought of you.
Show me that man in your arms, or love me, Thérèse! Failing these
alternatives, I can conceive but one other, and that is to kill myself and have
done with it. This is cowardly, stupid, the commonplace threat worn
threadbare by all despairing lovers; but is it my fault if there is a despair
which makes all those who undergo it utter the same shriek, and am I mad
because I happen to be a man like other men?
"Of what avail has been all that I have invented to protect myself from it,
and to render my poor personality as harmless as it wished to be free?
"Have you anything to reproach me for in my dealings with you,
Thérèse? Am I a conceited ass, a rake, I who prided myself upon making
myself stupid so as to give you confidence in my friendship? But why do
you wish that I should die without having loved, you who alone can teach
me what love is, and who know it well? You have a treasure in your heart,
and you smile as you sit beside a poor devil who is dying of hunger and
thirst. You toss him a small coin from time to time; that means friendship in
your eyes; it is not even pity, for you must know that the drop of water
increases the thirst.
"And why do you not love me? You may have loved some one who was
a worse man than I. I am not worth much, to be sure, but I love you, and is
not that everything?
"You will not believe it; you will say again that I am mistaken, as you said
before! No, you cannot say so, unless you lie to God and to yourself. You
see that my torment gets the better of me, and that I am making an absurd
declaration, I who dread nothing on earth so much as being laughed at by
you!
"Thérèse, do not think me corrupt. You know very well that the bottom of
my heart has never been sullied, and that from the abysses into which I
have cast myself I have always, in spite of myself, raised my voice to
Heaven. You know that with you I am as chaste as an infant, and you have
not feared to take my head in your hands sometimes, as if you were about
to kiss me on the forehead. And you would say: 'Bad head! you deserve to
be broken.'—And yet, instead of crushing it like the head of a snake, you
would try to make the pure and burning breath of your mind penetrate it.
Ah! well, you have succeeded only too well; and, now that you have kindled
the fire on the altar, you turn away and say to me: 'Place it in somebody's
else care! Marry; love some sweet, devoted, lovely girl; have children, be
ambitious for them, live a virtuous life, have a happy home—have
everything, except me!'
"But, Thérèse, it is you whom I love passionately, not myself. Since I
have known you, you have been striving to make me believe in happiness
and to cultivate my taste for it. It is not your fault if I have not become as
selfish as a spoiled child. But I am worthy of a better fate. I do not ask if
your love would mean happiness to me. I simply know that it would be life,
and that, good or bad, it is that life, or death, that I must have."
IV
Thérèse was deeply distressed by this letter. She was, as it were, struck by
lightning. Her love bore so little resemblance to Laurent's, that she fancied that
she did not love him with love, especially when she reread the expressions used
by him. There was no delirium in Thérèse's heart, or, if there were, it had
entered there, drop by drop, so slowly that she did not notice it, and believed
that she was as thoroughly mistress of herself as on the first day. The word
passion offended her.
"Passions! I?" she said to herself. "In Heaven's name, does he think that I
don't know what passion is, and that I want any more of that poisoned draught?
What have I done to him, I who have given him so much affection and thought,
that he should propose to me, by way of thanks, despair, madness, and death?
—After all," she thought, "it is not his fault, poor creature! He doesn't know what
he wants, nor what he asks for. He seeks love like the philosopher's stone, in
which one strives all the more earnestly to believe because he cannot grasp it.
He thinks that I have it, and that I amuse myself by refusing to give it to him!
There is always a touch of frenzy in whatever he thinks. How can I appease
him, and turn him aside from a fancy that is rapidly making him unhappy?
"It is my fault; he has some right to say so. While trying to wean him from
debauchery, I familiarized him too much with a virtuous attachment; but he is a
man, and he deems our affection incomplete. Why did he deceive me? why did
he make me believe that he was tranquil when he was with me? What shall I do
to repair the idiocy of my inexperience? I have not been enough of a woman in
the matter of presumption. I did not know that a woman, however lukewarm and
weary of life she may be, can still disturb a man's brain. I ought to have believed
that I was fascinating and dangerous, as he once told me, and to have guessed
that he afterward contradicted himself on that point only to quiet my fears. So it
is a misfortune, then,—for it certainly cannot be a sin,—not to have the instinct
of coquetry?"
Thereupon, Thérèse, searching her memory, remembered that she had
instinctively been reserved and distrustful to defend herself from the desires of
other men who were not attractive to her; with Laurent she felt no such instinct,
because she esteemed him in his friendship for her, because she could not
believe that he would seek to deceive her, and also, it must be admitted,
because she cared more for him than for any other. Alone in her studio, she
paced the floor, oppressed by a painful sense of discomfort, sometimes
glancing at the fatal letter which she had placed on a table as if she knew not
what to do with it and could neither decide to read it again nor to destroy it, and
sometimes looking at her unfinished work on the easel. She was working with
enthusiasm and pleasure when they brought her that letter, that is to say, that
doubt, that anxiety, that amazement, and that dread. It was like a mirage which
caused all the spectres of her former miseries to reappear upon her unclouded,
peaceful horizon. Every word written on that paper was like a hymn of death
which she had heard in the past, a prophecy of fresh misfortunes to come.
She tried to recover her serenity by resuming her painting. That was her
great remedy for all the petty ills of external life; but it was powerless that day;
the terror which that passion aroused in her assailed her in the purest and most
private sanctuary of her present life.
"Two pleasures disturbed or destroyed," she said to herself, throwing away
her brush and taking up the letter: "work and friendship."
She passed the rest of the day without making up her mind to anything. But
one point was perfectly clear in her mind, the determination to say no; but she
proposed that it should be no in reality, and was not bent upon announcing her
decision in hot haste, with the timorous abruptness of the woman who is afraid
of succumbing unless she makes haste to barricade the door. How to say that
no from which there should be no appeal, which should leave no hope behind,
yet should not be like a red-hot iron on the sweet memory of friendship, was a
hard and bitter problem for her to solve. That memory was her own love; when
one has the body of a dearly loved one to bury, one cannot decide, without bitter
sorrow, to place a white cloth over the face and bestow the body in the common
grave. You would fain embalm it in a grave of its own, which you could look
upon from time to time, praying for the soul of him whose dust it contains.
Darkness came upon her before she had hit upon any expedient for denying
herself without inflicting too much pain. Catherine, seeing that she dined with
little appetite, asked her anxiously if she were ill.
"No," she replied, "I have something on my mind."
"Ah! you work too hard," said the good old woman, "you do not think about
taking care of yourself."
Thérèse raised her finger; it was a gesture with which Catherine was familiar,
and which signified: "Don't speak of that."
The hour at which Thérèse received her few friends had, for some time past,
been taken advantage of by Laurent only. Although the door stood open for
whoever chose to come, he alone came, whether because the others were
away from Paris—it was the season for going into the country or remaining
there—or because they had detected in Thérèse a certain preoccupation, an
involuntary and poorly concealed desire to talk exclusively with Monsieur de
Fauvel.
Laurent usually arrived at eight, and Thérèse said to herself as she glanced
at the clock:
"I did not answer his letter; he won't come to-day."
There was a horrible void in her heart as she added:
"He must never come again."
How was she to pass that endless evening, which she was accustomed to
pass in conversation with her young friend, while she worked at some little
sketch or some fancy-work, and he smoked his cigar, half-reclining lazily on the
cushions of the couch? It occurred to her to escape the impending ennui by
calling upon a friend in Faubourg Saint-Germain, with whom she sometimes
went to the play; but her friend always retired early, and it would be too late
when she arrived. It was such a long way, and the cabs moved so slowly! Then,
too, she would have to dress, and Thérèse, who lived in slippers, like all artists
who work with enthusiasm and cannot bear to be incommoded by their clothes,
was very indolent in the matter of arraying herself in visiting costume. Suppose
she should put on a veil and a shawl, send for a cab, and drive slowly through
the deserted avenues of the Bois de Boulogne? Thérèse had sometimes taken
such a drive with Laurent, when the evening heat was so stifling that they
sought a breath of fresh air under the trees. Such excursions with anybody else
would have compromised her seriously; but Laurent guarded religiously the
secret of her confidence, and they both took delight in the unconventionality of
those mysterious tête-à-têtes, which concealed no mystery. She remembered
them as if they were already far away, and said to herself, sighing at the thought
that they would never return:
"Those were happy times! They can never come again for him who suffers,
or for me who am no longer in ignorance of it."
At nine o'clock, she at last attempted to answer Laurent's letter, but a peal of
the bell made her heart beat fast. It was he! She rose to tell Catherine to say
that she had gone out. Catherine returned: it was only a letter from him.
Thérèse had an involuntary thrill of regret that it was not himself.
There were only these few words in the letter:
"Adieu, Thérèse! you do not love me, and I love you like a child!"
These two lines made Thérèse tremble from head to foot. The only passion
that she had never striven to extinguish in her heart was maternal love. That
wound, although apparently healed, was still bleeding like unsatisfied love.
"Like a child!" she repeated, crumpling the letter in her quivering hands. "He
loves me like a child! Mon Dieu! what does he mean by that? does he know how
he hurts me? Adieu! My boy had learned to say adieu! but he did not say it
when they carried him away. I should have heard him! and I shall never hear
him again!"
Thérèse was overexcited, and, as her emotion seized upon the most painful
of pretexts for manifesting itself, she burst into tears.
"Did you call me?" said Catherine, entering the room. "Mon Dieu! what is the
matter? You are weeping as you used to in the old days!"
"Nothing, nothing, leave me," replied Thérèse. "If any one comes to see me,
say that I have gone to the theatre. I want to be alone. I am ill."
Catherine left the room, but went into the garden. She had seen Laurent
creeping stealthily along the hedge.
"Do not sulk like this," she said. "I don't know why my mistress is crying; but
it must be your fault, you make her unhappy. She doesn't want to see you. Go,
ask her pardon!"
Catherine, despite all her respect for Thérèse and devotion to her, was
satisfied that Laurent was her lover.
"She is weeping?" he cried. "Oh! mon Dieu! why is she weeping?"
And he rushed across the little garden and threw himself at Thérèse's feet,
who was sobbing in the small salon, with her face in her hands.
Laurent would have been overjoyed to see her thus if he had been the rake
that he sometimes sought to appear; but in reality he was wonderfully tender-
hearted, and Thérèse possessed the secret power of arousing his real nature.
The tears with which her face was bathed caused him genuine and profound
grief. On his knees he implored her to forget his madness, and to restore him to
reason by her gentleness and good sense.
"I wish only what you wish," he said, "and since you weep for our dead and
gone friendship, I swear that I will bring it to life again rather than cause you
fresh sorrow. But let us be frank with each other, my sweet, kind Thérèse, my
dearest sister, for I no longer feel the strength to deceive you! do you summon
the courage to accept my love as a deplorable discovery that you have made,
and as a disease of which you propose to cure me by patience and pity. I will do
my utmost in that direction, I swear to you! I will not ask you for so much as a
kiss, and I think that will not cost me so much as you might fear, for I do not yet
know whether my senses are involved in all this. No, really I do not think it. How
could it be so, after the life I have led and am at liberty to lead still? What I feel
is a thirst of the heart; why should it frighten you? Give me a little of your heart,
and take all of mine. Consent to be loved by me, and do not tell me again that it
is an insult to you, for what drives me to despair is to see that you despise me
too much to permit me, even in a dream, to aspire to you. That lowers me so in
my own eyes that it makes me long to kill the miserable devil who so offends
your moral sense. Raise me, rather, from the slough into which I have fallen,
and bid me expiate my evil life and become worthy of you. Yes, leave me a ray
of hope! however faint it may be, it will make another man of me. You will see,
you will see, Thérèse! The mere idea of striving to seem a better man to you
gives me strength already; I can feel it! do not take it from me. What will become
of me if you spurn me? I shall descend again all the steps I have climbed up
since I have known you. All the fruit of our sacred friendship will be lost, so far
as I am concerned. You will have tried to cure a sick man, and will have killed
him instead! And then, too, will you yourself, who are so noble and so good, be
content with your work? will you not reproach yourself for not having conducted
it to a better end? Be to me a Sister of Charity, who does not confine herself to
dressing the hurts of a wounded man, but who strives to reconcile his soul with
Heaven. Come, Thérèse, do not withdraw your faithful hands from mine, do not
turn away your face, so lovely in sorrow. I will not leave your feet until you have
at least forgiven me for loving you, if you have not authorized me to love you!"
Thérèse could but accept this effusion as serious, for Laurent was perfectly
honest. To hold him off with distrust would have been equivalent to an avowal of
the too warm affection which she had for him; a woman who shows fear is
already vanquished. So she assumed a brave face, and perhaps it was not
mere affectation, for she believed that her strength was still equal to the task.
Indeed, she was not ill-advised by her very weakness.
To break with him at that moment would have been to arouse terrible
emotions which it was much better to appease, reserving the right to relax the
bond gently, with skill and prudence. That might be a matter of some days.
Laurent was so impressionable, and rushed so abruptly from one extreme to the
other!
So they both calmed down, assisting each other to forget the storm, and
even exerting themselves to laugh at it, in order to afford each other mutual
encouragement touching the future; but, do what they would, their position was
essentially changed, and their intimacy had taken a giant's stride. The fear of
losing each other had brought them nearer together, and, while vowing that
there had been no change in their friendship, there was in all their words and
thoughts a sort of languor of the heart, a sort of rapturous fatigue, which was in
effect the enfranchisement of love!
Catherine, when she brought the tea, put them completely at their ease by
her artless, maternal solicitude.
"You would do much better," she said to Thérèse, "to eat the wing of a
chicken instead of making a hole in your stomach with this tea!—Do you know,"
she said to Laurent, pointing at her mistress, "that she never touched her
dinner?"
"Then let her sup at once!" cried Laurent. "Don't say no, Thérèse; you must
do it! What in Heaven's name would become of me if you should be taken sick?"
And as Thérèse refused to eat, for she really was not hungry, he declared, at
a sign from Catherine urging him to insist, that he was hungry himself; and that
was quite true, for he had forgotten to dine. Thereupon Thérèse was delighted
to give him some supper; and they ate together for the first time, which was no
trivial occurrence in Thérèse's lonely and modest life. To eat together is one of
the greatest promoters of intimacy. It is the satisfaction in common of a material
necessity of existence, and if you seek a loftier meaning in it, it is a communion,
as the word indicates.
Laurent, whose ideas naturally took a poetic turn, even in the midst of
merriment, laughingly compared himself to the Prodigal Son, for whom
Catherine made haste to kill the fatted calf. This fatted calf, which presented
itself in the guise of a meagre chicken, naturally added to the gaiety of the two
friends. It was so little for the young man's appetite to feed upon, that Thérèse
was distressed. The quarter was sadly lacking in resources, and Laurent
insisted that Catherine should not put herself out for him. In the depths of a
closet they unearthed a huge jar of guava jelly. It was a present from Palmer,
which Thérèse had forgotten to open, but which Laurent opened and attacked
with great zest, talking meanwhile with the utmost warmth of the excellent Dick,
of whom he had been foolish enough to be jealous, and whom he should love
thenceforth with all his heart.
"You see, Thérèse," he said, "how unjust disappointment makes us! Believe
me, children should be spoiled. Only those who are treated gently, turn out well.
So give me plenty of guavas, now and always! Harshness is not simply a bitter
gall, it is a deadly poison!"
When the tea arrived, Laurent discovered that he had devoured the chicken
like a selfish glutton, and that Thérèse had eaten nothing at all. He rebuked
himself for his heedlessness and confessed it; then he dismissed Catherine,
and insisted upon making the tea himself and waiting on Thérèse. It was the
first time in his life that he had ever waited on any one, and he found therein an
exquisite enjoyment which he recognized with ingenuous surprise.
THEIR FIRST SUPPER.
"Now", said Laurent to Thérèse, as he
knelt to offer her her cup, "I can understand
how one can be a servant and enjoy his
profession."
"Now," he said to Thérèse, as he knelt to offer her her cup, "I can understand
how one can be a servant and enjoy his profession."
From certain people, even the most trivial attentions have an extraordinary
value. There was in Laurent's manners, and even in his attitudes, a certain
stiffness which he never laid aside, even with society women. He offered them
attentions with the ceremonious coldness of the most rigid etiquette. With
Thérèse, who did the honors of her little home like the excellent woman and
good-humored artist that she was, he had always been cared for and coddled
without having to reciprocate. He would have shown a lack of taste and of tact
in assuming to act as the man of the house. Suddenly, as a result of those tears
and mutual outpourings of the heart, he found himself, without in the least
understanding how it happened, invested with rights which did not belong to
him, but which he seized upon as if by inspiration, without opposition on the part
of the surprised and deeply moved Thérèse. It seemed to him that he was under
his own roof, and that he had won the privilege of looking after the mistress of
the house, like a loving brother or an old friend. And Thérèse, without a thought
of the danger of this taking possession, watched him with wide-open, wondering
eyes, asking herself if she had not been radically mistaken hitherto in regarding
that affectionate and devoted child as a reserved and gloomy man.
However, Thérèse reflected during the night; but, in the morning, Laurent,
who, although he had no premeditated plan, did not propose to give her time to
breathe,—for he had ceased to breathe himself,—sent her magnificent flowers,
rare sweetmeats, and a note so loving, so gentle and respectful, that she could
not fail to be touched by it. He said that he was the happiest of men, that he
desired nothing more except her forgiveness, and that, as soon as he had
obtained that, he should be the king of the world. He would accept any
deprivation, any harsh decree, provided only that he was not forbidden to see
and talk with his friend. That alone would be beyond his strength; all the rest
was as nothing. He was well aware that Thérèse could have no love for him;
which fact did not deter him from saying, ten lines lower down: "Is not our
sanctified love indissoluble?"
And stating thus the pros and the con's, the false and the true, a hundred
times a day, with a candor by which he was unquestionably deceived himself,
encompassing Thérèse with delicate attentions, striving with all his heart to give
her confidence in the chastity of their relations, and at every instant talking to
her in a lofty strain of his adoration for her, seeking to divert her when she was
disturbed in mind, to cheer her when she was sad, to melt her toward himself
when she was stern, he led her insensibly to the point where she had no other
will and no other existence than his.
Nothing is so perilous as those intimacies in which the parties have
exchanged a promise not to attack each other, when neither of the two inspires
a secret physical repulsion in the other. Artists, because of their independent life
and the nature of their occupations, which oblige them often to depart from
social conventions, are more exposed to these perils than they who live by rule
and whose imaginations are less active. We should forgive them, therefore, for
more sudden impressions and more feverish impulses. Public opinion realizes
this duty, for it is generally more indulgent to those who go astray, perforce, in
the tempest than those whose lives are passed in a flat calm. And then the
world demands from artists the fire of inspiration, and that fire, which overflows
for the enjoyment and enthusiasm of the public, must inevitably consume
themselves in time. Then we pity them; and the honest bourgeois, returning to
his family at night, after learning of their misfortunes and downfall, says to his
excellent and gentle helpmeet:
"You remember that poor girl who sang so well? she is dead of a broken
heart. And that famous poet who wrote such fine verses has killed himself. It's a
great pity, wife. All those people end badly. We, the simple creatures, are the
happy ones."
And the honest bourgeois is right.
Thérèse had lived a long while, however, if not as an honest bourgeois,—for
one must have a family for that, and God had denied her a family,—at all events
as a hard-working young woman, working from early morning, and not regaling
herself with pleasure or sloth at the end of her day's work. She constantly
aspired to the joys of a regular, domestic life; she loved order, and, far from
displaying the childish contempt which certain artists lavished upon what they
called, in those days, the grocer class, she bitterly regretted that she had not
married in that safe and modest social circle where she would have found
affection and security instead of talent and renown. But we do not choose our
own destinies, for fools and ambitious mortals are not the only imprudent wights
whom destiny overwhelms.
Thérèse had no weakness for Laurent in the ironical, libertine sense in which
that word is commonly used in love. It was by an effort of her will that she said
to him, after many nights of painful meditation:
"I wish what you wish, because we have reached the point where the sin still
to be committed is the inevitable reparation of a succession of sins already
committed. I have been culpable toward you, in not having had the selfish
prudence to fly from you; it is better that I should be culpable toward myself by
remaining your companion and your consolation, at the price of my peace of
mind and my pride.—Listen," she added, grasping his hand with all the strength
of which she was capable, "do not withdraw this hand, and, whatever happens,
always retain enough honor and courage to remember that before being your
mistress I was your friend. I said it to myself on the first day of your passion: we
loved each other too well thus, not to love each other less well under other
circumstances; but that happiness could not last for me, because you no longer
share it, and pain has taken the upper hand in that liaison, in which for you pain
and pleasure are mingled. I simply ask you, if you become weary of my love, as
you become weary of my friendship, to remember that it is not an instant of
frenzy that has driven me into your arms, but an impulse of my heart and a
more loving and lasting sentiment than voluptuous intoxication. I am not
superior to other women, and I do not assume the right to deem myself
invulnerable; but I love you so ardently and so purely that I should never have
transgressed with you, if you could have been saved by my strength. After I had
believed that strength was beneficial to you, that it would teach you to discover
your own strength and to purge yourself of an evil past, I found that you were
convinced of the contrary, so convinced that to-day the contrary has actually
happened: you are becoming bitter, and it seems that, if I resist, you are ready
to hate me and to return to your life of debauchery, blaspheming even our poor
friendship. And so I offer to God the sacrifice of my life for you. If I am destined
to suffer because of your nature or your past, so be it. I shall be amply repaid if I
rescue you from the suicide you were in a fair way to commit when I first knew
you. If I do not succeed, I shall at least have made the trial, and God, who
knows how sincere my devotion is, will pardon me if it is of no avail!"
In the early days of this union, Laurent's enthusiasm, gratitude, and faith
were admirable to see. He rose superior to himself, he had outbursts of religious
fervor, he blessed his dear mistress for having made known to him at last the
true, chaste, and noble love of which he had dreamed so much, and of which he
had thought that he was to be deprived forever by his own fault. She dipped him
anew, he said, in the waters of his baptism, she wiped out even the memory of
his evil days. It was adoration, worship, ecstatic contemplation.
Thérèse ingenuously believed in him. She abandoned herself to the joy of
having caused all that happiness and restored all that grandeur of soul to one of
God's elect. She forgot all her apprehensions, or smiled at them as meaningless
dreams which she had mistaken for arguments. They laughed at them together;
they reproached themselves for having misunderstood each other and for not
having thrown themselves on each other's neck the very first day, they were so
perfectly adapted to understand, appreciate, and cherish each other. There was
no more talk of prudence, no more sermons. Thérèse had grown ten years
younger. She was a child, more childish than Laurent himself; she devoted all
her energies to the task of arranging his existence so that he would not feel the
fold of a rose-leaf.
Poor Thérèse! Her intoxication did not last eight whole days.
Whence comes that terrible chastisement inflicted on those who have
abused the forces of youth, a chastisement which consists in making them
incapable of appreciating the joys of a harmonious and logical existence? Is the
young man a very great criminal who, being launched in the world without a
curb and with boundless aspirations, deems himself capable of exterminating all
the phantoms that pass, of mastering all the pleasures that beckon to him? Is
his sin anything else than ignorance, and could he have learned in his cradle
that life should be a constant battle with one's self? There are some of these
young men who are really to be pitied, and whom it is difficult to condemn, some
who may have been without a guide, a careful mother, a judicious friend, a
sincere first mistress. Vertigo has seized them at the outset; corruption has
hurled itself upon them as upon its lawful prey, to make brutes of those who
have more senses than heart, to make madmen of those who struggle, as
Laurent did, between the mire of reality and the ideal of their dreams.
That is what Thérèse said to herself in order to keep on loving that suffering
soul, and why she endured the outrages we are about to describe.
The seventh day of their happiness was irrevocably the last. That ill-fated
figure was never absent from Thérèse's mind. Fortuitous circumstances
combined to prolong that eternity of joy throughout a whole week; no one with
whom she was intimate had come to see Thérèse, she had no work that was
urgent; Laurent promised to set to work afresh as soon as he could resume
possession of his studio, then in the hands of workmen who were making
certain repairs. The heat in Paris was most oppressive; he proposed to Thérèse
that they should pass forty-eight hours in the country, in the woods. It was the
seventh day.
They set off by boat, and arrived at night-fall at a hotel. After dinner, they
went out to ride in the forest in the lovely moonlight. They hired horses and a
guide, who soon wearied them by his boastful loquacity. They had ridden about
two leagues when they came to the foot of a cliff with which Laurent was
familiar. He proposed to dismiss the horses and the guide, and to return on foot,
even if it were a little late.
"I am sure," said Thérèse, "I don't know why we should not pass the whole
night in the forest. There are no wolves or thieves here. Let us stay as long as
you choose, and never go back, if you say so."
They were left alone; and thereupon a strange, almost impossible scene
occurred, which we must describe just as it happened. They had climbed to the
top of the cliff, and were sitting on the thick moss, which was burned and
withered by the intense heat. Laurent was gazing at the magnificent spectacle
of the sky, where the moon dimmed the brilliancy of the stars. Only two or three
of the largest could be distinguished in the zenith. Laurent, lying on his back,
gazed at them.
"I would like to know," he said, "the name of this one almost over my head; it
seems to be looking at me."
"That is Vega," said Thérèse.
"So you know the names of all the stars, do you, my learned lady?"
"Of almost all. It isn't difficult, and in fifteen minutes you shall know as many
as I do, if you choose."
"No, thanks; I much prefer not to know them; I prefer to give them names of
my own."
"And you are right."
"I prefer to roam at random among those lines up yonder, and arrange
combinations of groups according to my own ideas, rather than walk according
to the whims of other people. After all, perhaps I am wrong, Thérèse! You prefer
the beaten paths, don't you?"
"They are softer for tender feet. I haven't seven-league boots like you!"
"What sarcasm! you know very well that you are a stronger and better walker
than I!"
"That is easily understood: it is because I have no wings to fly."
"See to it that you don't get hold of any and leave me behind! But let us not
talk about parting: that word would make the heavens weep!"
"What! who has any idea of such a thing? Don't say that ghastly word again!"
"No, no! let's not think of it!" he cried, springing suddenly to his feet.
"What's the matter, and where are you going?" she said.
"I don't know," he replied. "Ah! by the way——There's an extraordinary echo
here, and the last time I came here with little—you don't care to know her name,
do you? I enjoyed listening to it here, while she sang on yonder little hillock, just
opposite us."
Thérèse made no reply. He saw that this unseasonable evocation of one of
his undesirable acquaintances was not a very delicate contribution to the joys of
a romantic midnight expedition with the queen of his heart. Why had it come into
his head? how was it that the name of some foolish virgin or other had come to
his lips? He was mortified by his blunder; but, instead of ingenuously blaming
himself for it and wooing oblivion by torrents of loving words which he was quite
capable of pouring forth when passion inspired his heart, he determined to
brazen it out, and asked Thérèse if she would sing for him.
"I could not do it," she replied, gently. "It is a long while since I have ridden,
and it has made me a little oppressed."
"If it is only a little, make an effort, Thérèse; it will give me so much
pleasure!"
Thérèse was too proud to be angry, she was only grieved. She turned her
face away, and pretended to cough.
"Well, well," said he, laughingly, "you are only a poor weak woman! And then
you don't believe in my echo, I can see that. I propose that you shall hear it.
Stay here. I will climb up to the top of the hill. You are not afraid to remain alone
for five minutes, I trust?"
"No," replied Thérèse, "I am not at all afraid."
To climb to the other height, he had to go down into the little ravine which
separated it from that on which they were; but the ravine was deeper than it
seemed. When Laurent, having gone down half-way, saw how far he still had to
go, he stopped, reluctant to leave Thérèse alone so long, and called to her,
asking if she had called him.
"No, indeed I did not!" she shouted, not wishing to thwart his caprice.
It is impossible to describe what took place in Laurent's brain; he took that
indeed I did not for a rebuke, and continued to descend, but less quickly, and
musing as he walked.
"I have wounded her," he said, "and now she is sulky, as in the days when
we played at being brother and sister. Is she going to continue to have these
moody fits, now that she is my mistress? But why did I wound her? I was wrong,
certainly, but it was unintentional. It is impossible that some scrap of my past
should not come to my mind now and then. Is it to be an insult to her and a
mortification to me every time? What does my past matter to her, since she has
accepted me as I am? And yet I was wrong! yes, I was wrong; but will she never
happen to mention the wretch whom she loved and thought she had married? In
spite of herself, Thérèse when she is with me will remember the days she lived
without me, and shall I twist it into a crime?"
He instantly answered his own question:
"Oh! yes, it would be intolerable! So I have done very wrong, and I ought to
have asked her pardon at once."
But he had already reached that stage of mental fatigue when the mind is
sated with enthusiasm, and when the weak and shrinking creature that every
one of us is, to a greater or less extent, feels that he must resume possession of
himself.
"Must humble myself again, promise again, persuade again, shed tears
again?" he said to himself. "Great God! can't she be happy and trustful for a
single week? It's my fault, I agree, but it's hers still more for making so much out
of so little, and spoiling this lovely poetic night which I had planned to pass with
her in one of the loveliest spots on earth. I have been here before with rakes
and wantons, it is true; but to what corner of the outskirts of Paris could I have
taken her where I was not likely to run against some such unpleasant
reminiscence? Surely they can hardly be said to enchant me, and it is almost
cruel to reproach me with them."
As he replied thus in his heart to the reproaches with which Thérèse was
probably upbraiding him in hers, he reached the bottom of the ravine, where he
felt disturbed and fatigued as after a quarrel, and threw himself on the grass in a
fit of annoyance and weariness. For seven whole days he had not belonged to
himself; he was conscious of a longing to reconquer his liberty, and to fancy
himself alone and unsubdued for a moment.
Thérèse, for her part, was heart-broken and terrified at the same time. Why
had the word parting been suddenly hurled at her like a shrill cry amid that
tranquil atmosphere that they were breathing together? what was the occasion
of it? how had she provoked it? She tried in vain to determine. Laurent himself
could not have explained it to her. All that had followed was grossly cruel, and
how angry he must have been, that man of exquisite breeding, to have said it!
But what was the cause of his anger? Had he a serpent within him that gnawed
his heart and extorted from him wild and blasphemous words?
She had followed him with her eyes down the slope until he had entered the
dark shadow of the ravine. Then she saw him no more, and was surprised at
the time that elapsed before he appeared on the slope of the other hill. She was
frightened at last—he might have fallen over some precipice. In vain did her
eager glances question the grass-grown depths, bristling with huge, dark rocks.
She rose, intending to call to him, when an indescribable cry of distress reached
her ears, a hoarse, ghastly, desperate shriek, which made her hair stand on
end.
She darted away like an arrow in the direction of the voice. If there had really
been a precipice, she would have rushed over the edge without reflection; but
there was only a steep incline, where she slipped several times on the moss
and tore her dress on the bushes. Nothing stopped her; she reached Laurent's
side, how, she knew not, and found him on his feet, with haggard eyes and
trembling convulsively.
"Ah! here you are," he said, grasping her arm; "you did well to come! I should
have died!"
And, like Don Juan after the reply of the statue, he added in a sharp, abrupt
tone: "Let us go away from here!"
He led her to the road, walking at random, and unable to tell what had
happened to him.
He became calmer at last, after ten or fifteen minutes, and sat down beside
her in a clearing. They had no idea where they were; the ground was strewn
with flat rocks which resembled tombs, and among them grew juniper-trees,
which one might well have mistaken at night for cypresses.
"Mon Dieu!" said Laurent, suddenly, "are we in a cemetery? Why did you
bring me here?"
"It is simply a tract of untilled land," she replied. "We passed through many
similar ones this evening. If you don't like it, let's not stop here, but go back
under the large trees."
"No, let us stay here," he rejoined. "Since chance or destiny brings these
thoughts of death into my mind, I may as well face them and exhaust their
horror. They have their charm, like everything else, have they not, Thérèse?
Everything that moves the imagination powerfully, affords us enjoyment, more or
less painful. When a head is to fall on the scaffold, the multitude goes to see,
and it is altogether natural. We cannot live upon mild emotions alone; we need
terrible emotions to make us realize the intensity of life."
He spoke thus, as if at random, for some moments. Thérèse dared not
question him, and strove to divert his thoughts; she saw that he had had an
attack of delirium. At last, he recovered sufficiently to desire and to be able to
describe it to her.
He had had an hallucination. As he lay on the grass, in the ravine, his brain
had become confused. He had heard the echo sing all by itself, and its song
was an obscene refrain. Then, as he raised himself on his elbow to seek an
explanation of the phenomenon, he had seen a man rush toward him over the
grass, a pale-faced man, with torn garments, and his hair flying in the wind.
"I saw him so plainly," he said, "that I had time to say to myself that he was
some belated citizen who had been surprised and pursued by robbers, and I
even looked for my cane to go to his assistance, but my cane was lost in the
grass, and the man was still running toward me. When he was close to me, I
saw that he was drunk and that no one was chasing him. As he passed me, he
cast a stupid, hideous glance at me, and made a fiendish grimace of hatred and
contempt. Then I was afraid, and threw myself face downward on the ground,
for that man—was myself!—Yes, it was my ghost, Thérèse. Don't be frightened,
don't think me mad, it was a vision. I realized it when I found myself alone in the
darkness. I could not have distinguished the features of any human face, I had
seen that one only in my imagination; but how distinct, how ghastly, how horrible
it was. It was myself twenty years older, with features wasted by debauchery or
disease, wild eyes, a brutalized mouth, and, despite the total wiping out of my
vitality, there was enough energy remaining in that phantom to insult and defy
the creature that I am now. Thereupon, I said to myself: 'O my God! is that what
I shall be in my mature years?' This evening, there came into my mind the
memory of some miserable past experiences of mine, and I blurted them out
involuntarily; is it because I still bear within me that old man from whom I
believed that I was free? The spectre of debauchery will not release his victims,
and even in Thérèse's arms he will mock at me and cry: 'It is too late!'
"Then I rose to go back to you, my poor Thérèse. I intended to ask your
pardon for my vileness and to beg you to save me; but I don't know how many
minutes or centuries I should have turned round and round, unable to take a
step forward, if you had not come to me at last. I recognized you instantly,
Thérèse; I was not afraid of you, and I felt that I was saved."
It was difficult to determine, when Laurent talked thus, whether he was telling
of something that had really happened, or whether he had confused in his brain
an allegory born of his bitter reflections and a vision which he had seen
indistinctly in a sort of half-sleep. He swore, however, that he did not fall asleep
on the grass, and that he had not once lost consciousness of the place where
he was or of the passage of time; but even that was difficult to decide. Thérèse
had lost sight of him, and to her the time had seemed mortally long.
She asked him if he were subject to these hallucinations.
"Yes," he said, "when I am drunk; but I have been drunk with love only, in the
fortnight that you have been mine."
"The fortnight!" echoed Thérèse in amazement.
"No, less than that," he replied; "don't split hairs with me about dates; you
see that I have not my wits as yet. Let us walk, that will fix me all right."
"But you need rest; we must think of returning to the hotel."
"Well, what are we doing?"
"We are not going in the right direction; our backs are turned to our starting-
point."
"Do you want me to go back over that infernal rock?"
"No, but let us go to the right."
"That is just the wrong direction."
Thérèse insisted that she was not mistaken. Laurent refused to yield; he
even lost his temper, and spoke in an irritated tone, as if that were a subject for
quarrelling. Thérèse yielded at last, and followed where he chose to go. She
was completely crushed by emotion and sadness. Laurent had spoken to her in
a tone which she had never dreamed of assuming with Catherine, even when
the old servant angered her. She forgave him because she felt that he was ill;
but his state of painful excitement alarmed her so much the more.
Thanks to Laurent's obstinacy, they lost themselves in the forest, walked
about four hours, and did not return to the hotel until daybreak. The walking in
the fine, heavy sand of the forest is very fatiguing. Thérèse could hardly drag
herself along, and Laurent, revivified by that violent exercise, did not think of
slackening his pace out of regard for her. He walked ahead, constantly asserting
that he had found the right path, asking her from time to time if she were tired,
and never suspecting that, when she answered no, her purpose was to spare
him any regret for having caused the misadventure.
The next day, Laurent had forgotten all about it; he had suffered a severe
shock, however, from that strange attack, but it is a peculiarity of excessively
nervous temperaments that they recover from such shocks as if by magic.
Indeed, Thérèse had occasion to notice, that on the day following that
lamentable experience it was she who was utterly exhausted, while he seemed
to have acquired fresh strength.
She had not slept, expecting to find him suffering from some serious illness;
but he took a bath, and felt quite disposed to repeat the excursion. He seemed
to have forgotten how disastrous that nocturnal experience had been to the
honeymoon! The melancholy impression soon wore away, so far as Thérèse
was concerned. When they returned to Paris, she imagined that nothing had
changed between them; but that same evening Laurent chose to draw a
caricature of Thérèse and himself wandering through the forest by moonlight, he
with his wild, distraught look, she with her torn gown and her fatigue-stricken
body. Artists are so accustomed to make caricatures of one another, that
Thérèse was amused by this one; but, although she, too, had plenty of facility
and humor at the end of her pencil, she would not have caricatured Laurent for
anything in the world; and when she saw him sketch from a comical standpoint
that nocturnal scene which had so tortured her, she was deeply grieved. It
seemed to her that there are certain sorrows of the heart which can never have
a ridiculous side.
Laurent, instead of understanding her feeling, became still more satirical. He
wrote under his own figure: Lost in the forest and in his mistress's heart; and

Case Based Reasoning Research and Development 26th International Conference ICCBR 2018 Stockholm Sweden July 9 12 2018 Proceedings Michael T. Cox

Uploaded by

Copyright:

Available Formats

Case Based Reasoning Research and Development 26th International Conference ICCBR 2018 Stockholm Sweden July 9 12 2018 Proceedings Michael T. Cox

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Case Based Reasoning Research and Development 26th International Conference ICCBR 2018 Stockholm Sweden July 9 12 2018 Proceedings Michael T. Cox

Uploaded by

Copyright:

Available Formats

Download the full version of the textbook now at textbookfull.

Case Based Reasoning Research and Development

Explore and download more textbook at https://textbookfull.com

Case Based Reasoning Research and Development 28th

Case Based Reasoning Research and Development 22nd

Artificial Intelligence for Knowledge Management 6th IFIP

Oil Booms and Business Busts Why Resource Wealth Hurts

ODI from BRIC Countries Firm level Evidence Valeria Gattai

Regional Conference on Science Technology and Social

Snowden Official Motion Picture Edition Kieran Fitzgerald

Short-Circuits in AC and DC Systems: ANSI, IEEE, and IEC

Subseries of Lecture Notes in Computer Science

LNAI Series Editors

LNAI Founding Series Editor

Shahina Begum (Eds.)

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Library of Congress Control Number: 2018955569

LNCS Sublibrary: SL7 – Artiﬁcial Intelligence

© Springer Nature Switzerland AG 2018

a day-long workshop entitled “Automating the Construction of Case-Based Reason-

July 2018 Michael T. Cox

Yves Lepage Waseda University, Japan

Ahmed, Mobyen Uddin Kübler, Eric

Adaptive Goal Driven Autonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Main Technical Papers

Answering with Cases: A CBR Approach to Deep Learning. . . . . . . . . . . . . 15

CEC-Model: A New Competence Model for CBR Systems Based

The Case for Case Based Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Case Based Reasoning as a Model for Cognitive Artificial Intelligence . . . . . 62

Explainable Distributed Case-Based Support Systems:

Tangent Recognition and Anomaly Pruning to TRAP Off-Topic Questions

Combining Case-Based Reasoning with Complex Event Processing

AI-VT: An Example of CBR that Generates a Variety of Solutions

A Textual Recommender System for Clinical Data . . . . . . . . . . . . . . . . . . . 140

Harnessing Hundreds of Millions of Cases: Case-Based Prediction

Case Base Elicitation for a Context-Aware Recommender System . . . . . . . . . 170

The SECCO Ontology for the Retrieval and Generation

Exploration vs. Exploitation in Case-Base Maintenance:

Dynamic Detection of Radical Profiles in Social Networks Using Image

Segmentation of Kidneys Deformed by Nephroblastoma Using

FITsense: Employing Multi-modal Sensors in Smart Homes

Embedded Word Representations for Rich Indexing: A Case Study

Case-Based Data Masking for Software Test Management . . . . . . . . . . . . . . 281

A CBR Approach for Imitating Human Playing Style in Ms. Pac-Man

Perks of Being Lazy: Boosting Retrieval Performance . . . . . . . . . . . . . . . . . 309

Bayesian-Supported Retrieval in BNCreek: A Knowledge-Intensive

Personalised Human Activity Recognition Using Matching Networks . . . . . . 339

Why Did Naethan Pick Android over Apple? Exploiting Trade-offs

An Analysis of Case Representations for Marathon Race Prediction

Dynamic Case Bases and the Asymmetrical Weighted

Novel Object Discovery Using Case-Based Reasoning and Convolutional

Modelling Similarity for Comparing Physical Activity

Investigating Textual Case-Based XAI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431

Improving kNN for Human Activity Recognition with Privileged Learning

Considering Nutrients During the Generation of Recipes

An Effective Method for Identifying Unknown Unknowns

Special Track: Computational Analogy

On the Role of Similarity in Analogical Transfer . . . . . . . . . . . . . . . . . . . . 499

Predicting Preferences by Means of Analogical Proportions . . . . . . . . . . . . . 515