PDF Relaxed Abduction Robust Information Interpretation For Industrial Applications 1St Edition Thomas Hubauer Auth Ebook Full Chapter

Relaxed Abduction Robust Information
Interpretation for Industrial

Applications 1st Edition Thomas
Hubauer (Auth.)
Visit to download the full and correct content document:
https://textbookfull.com/product/relaxed-abduction-robust-information-interpretation-fo
r-industrial-applications-1st-edition-thomas-hubauer-auth/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Uncertainty Management for Robust Industrial Design in

Aeronautics Charles Hirsch
https://textbookfull.com/product/uncertainty-management-for-
robust-industrial-design-in-aeronautics-charles-hirsch/
Sensual Abduction Series Box Set Sensual Abduction 1 4

Wilson Amelia
https://textbookfull.com/product/sensual-abduction-series-box-
set-sensual-abduction-1-4-wilson-amelia/
Computer Information Systems and Industrial Management:

17th International Conference, CISIM 2018, Olomouc,
Czech Republic, September 27-29, 2018, Proceedings
Khalid Saeed
https://textbookfull.com/product/computer-information-systems-
and-industrial-management-17th-international-conference-
cisim-2018-olomouc-czech-republic-
september-27-29-2018-proceedings-khalid-saeed/
Robust Control: Theory and Applications Kang-Zhi Liu
https://textbookfull.com/product/robust-control-theory-and-
applications-kang-zhi-liu/
Mechatronics 2017 Ideas for Industrial Applications
Jerzy ■wider
https://textbookfull.com/product/mechatronics-2017-ideas-for-
industrial-applications-jerzy-swider/
Industrial chemistry of oxides for emerging

applications First Edition Blanchart
https://textbookfull.com/product/industrial-chemistry-of-oxides-
for-emerging-applications-first-edition-blanchart/
Recombinant Microbes for Industrial and Agricultural

Applications First Edition Imanaka
https://textbookfull.com/product/recombinant-microbes-for-
industrial-and-agricultural-applications-first-edition-imanaka/
Industrial Enzyme Applications 1st Edition Andreas

Vogel
https://textbookfull.com/product/industrial-enzyme-
applications-1st-edition-andreas-vogel/
Machine Learning Algorithms for Industrial Applications

Santosh Kumar Das
https://textbookfull.com/product/machine-learning-algorithms-for-
industrial-applications-santosh-kumar-das/
Thomas Hubauer
Relaxed Abduction
Robust Information Interpretation
for Industrial Applications
Relaxed Abduction
Thomas Hubauer
Relaxed Abduction
Robust Information Interpretation
for Industrial Applications
Thomas Hubauer
München, Deutschland
Dissertation at the University of Lübeck, 2015
ISBN 978-3-658-14406-7 ISBN 978-3-658-14407-4 (eBook)

DOI 10.1007/978-3-658-14407-4
Library of Congress Control Number: 2016942010
Springer Vieweg
© Springer Fachmedien Wiesbaden 2016
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.
Printed on acid-free paper
This Springer Vieweg imprint is published by Springer Nature

The registered company is Springer Fachmedien Wiesbaden GmbH
I owe thanks to a number of people whose continued support made this
thesis possible. First and foremost, I would like to express my gratefulness
to my family for their encouragement and backing during these years. My
thanks extend to Prof. Dr. Ralf Möller for supervising and teaching me as
an external PhD candidate over many years, and to Prof. Dr. Till Tantau
for his readiness to review my thesis. Last in order but definitely not least
in importance, my gratitude goes to Dr. Özgür Özcep, Dr. Steffen
Lamparter, and Dr. Stephan Grimm for their criticism, suggestions, and
continued encouragement.
Preface
Automated information interpretation is gaining momentum in industrial
applications. One of the major challenges in this context is the appropriate
treatment of incompleteness of both the observations about the world,
and the domain model formalizing it. This dissertation proposes a novel
approach called "Relaxed Abduction", which is able to provide reasonable
interpretations even if it would not be possible or extremely complex to
explain all observations made. The approach is based on the idea of treating
explanatory power and consilience of an interpretation bi-criterially instead
of mapping them onto a one-dimensional scale. Based on a formalization
of the proposed approach, the thesis investigates concrete instantiations
and their properties (particularly runtime complexity), and proposes an
extension which allows to handle incoming changes in the underlying data
incrementally. Two algorithms for solving relaxed abduction problems are
proposed (one generic and one EL-specific) and evaluated in a real-world
use case, confirming the theoretical results in practice. Additionally, to
close the gap between academic research and industrial applications, the
thesis proposes a methodology to structure diagnosis problems according to
ISO 13379 and express it by means of several description logic knowledge
bases. The dissertation results show that the proposed, flexible notion
of abduction is indeed novel, relevant for industry, and permits practical
usage. Future steps required to bring relaxed abduction into application are
identified.
Contents
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Scientific Contributions . . . . . . . . . . . . . . . . . . . . . 5
1.4 Dissemination Activities . . . . . . . . . . . . . . . . . . . . . 6
1.5 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Preliminaries 9
2.1 Ontologies and Description Logics . . . . . . . . . . . . . . . 9
2.2 Abductive Reasoning . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Standard Notions of Abduction . . . . . . . . . . . . . 12
2.2.2 Concept-based Notions of Abduction . . . . . . . . . . 15
2.2.3 A Critique of Logic-Based Abduction . . . . . . . . . 17
2.3 Orders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Hypergraphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3 Relaxed Abduction 27
3.1 Formalising Diagnostics using Description Logics . . . . . . . 27
3.1.1 A Conceptual Model for Diagnostics . . . . . . . . . . 28
3.1.2 Formalising Diagnostics in EL+ . . . . . . . . . . . . . 29
3.2 Formalization and Properties of Relaxed Abduction . . . . . 33
3.2.1 Bi-Criteriality of Information Interpretation . . . . . . 33
3.2.2 Relaxing Logic-Based Abduction . . . . . . . . . . . . 34
3.2.3 Instantiating Relaxed Abduction: Mapping to Dia-
gnostics . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.4 Instantiating Relaxed Abduction: Choices for A and
O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3 Solving Relaxed Abduction . . . . . . . . . . . . . . . . . . . 50
3.3.1 A Generic Solution . . . . . . . . . . . . . . . . . . . . 51
3.3.2 A Completion-based Algorithm for EL+ . . . . . . . . 55
3.3.3 Employing Pruning to Increase Efficiency . . . . . . . 74
3.3.4 Axiom Selection Strategies . . . . . . . . . . . . . . . 80
X Contents
3.4 Extensions to the Basic Setting . . . . . . . . . . . . . . . . . 83

3.4.1 Extending Completion-based Relaxed Abduction to
More Expressive DLs . . . . . . . . . . . . . . . . . . . 83
3.4.2 Incremental Relaxed Abduction . . . . . . . . . . . . . 93
3.5 Comparison to Alternative Approaches . . . . . . . . . . . . . 103
3.5.1 Information Interpretation and Diagnostics . . . . . . 103
3.5.2 Consequence-Driven Reasoning Methods . . . . . . . . 106
3.5.3 Truth Maintenance and Reasoning . . . . . . . . . . . 106
4 Case Studies and Evaluation 109

4.1 Implementation of the RAbIT System . . . . . . . . . . . . . . 109
4.2 Performance and Scalability . . . . . . . . . . . . . . . . . . . 110
4.3 Application Scenarios and Expert Feedback . . . . . . . . . . 116
4.3.1 Scenario: Turbine Diagnostics . . . . . . . . . . . . . . 116
4.3.2 Scenario: Debugging of Rule-Bases . . . . . . . . . . . 118
5 Conclusion 121
5.1 Discussion of Results . . . . . . . . . . . . . . . . . . . . . . . 121
5.1.1 Scientific Contributions . . . . . . . . . . . . . . . . . 121
5.1.2 Practical Applications . . . . . . . . . . . . . . . . . . 122
5.2 Directions for Future Work . . . . . . . . . . . . . . . . . . . 123
List of Figures
2.1 Pareto-optimality and Dominance . . . . . . . . . . . . . . . . 22
3.1 A conceptual model for diagnostics . . . . . . . . . . . . . . . 28

3.2 ER representation of core entities in diagnostics . . . . . . . . 30
3.3 A hierarchy of diagnostic ontologies . . . . . . . . . . . . . . . 31
3.4 Normalisation rules for EL+ . . . . . . . . . . . . . . . . . . 56
3.5 Completion rules for classification in EL+ . . . . . . . . . . . 57
3.6 Completion rules for relaxed abduction in EL+ . . . . . . . . 63
3.7 Completion rules for classification in EL++ . . . . . . . . . . 85
4.1 Relative runtime of RAPsolve-EL+ vs. RAPsolve-Generic . . 113

4.2 Runtime of RAPsolve-EL+ vs. scaling factor xA . . . . . . . . 114
4.3 Runtime of RAPsolve-EL+ vs. scaling factor xA . . . . . . . . 115
List of Tables
2.1 EL+ syntax and semantics . . . . . . . . . . . . . . . . . . . . 12
3.1 Relation between type of knowledge modelling, reasoning

method, and application . . . . . . . . . . . . . . . . . . . . . 41
3.2 EL++ syntax and semantics . . . . . . . . . . . . . . . . . . . 84
3.3 Potential for improvements in efficiency by reusing information
in RAPsolve-Generic and RAPsolve-EL++ . . . . . . . . . . . 102
3.4 Comparison of related approaches . . . . . . . . . . . . . . . . 108
Other Listings
List of Examples
1.1 A simple diagnostics scenario . . . . . . . . . . . . . . . . . . . . 4
2.1 Diagnostics over incomplete domain formalisations . . . . . . . . 17
3.1 Diagnostics over incomplete domain formalisations (cont.) . . . . 43
3.2 Subset- and simple entailment-based RAP are incomparable . . 45
3.3 Comparing nclusion- and cardinality-based orders . . . . . . . . . 49
3.4 Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.5 Effects of Pruning . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.6 Incrementality in diagnostics (updating O) . . . . . . . . . . . . 94
3.7 Incremenatlity in diagnostics (updating T ) . . . . . . . . . . . . 95
3.8 Incremenatlity in diagnostics (updating A) . . . . . . . . . . . . 95
3.9 Incremenatlity in diagnostics (axiom migration) . . . . . . . . . . 96
List of Definitions
2.1 Logically independent axiom set . . . . . . . . . . . . . . . . . . 10
2.2 Abduction problem . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Preferential abduction problem . . . . . . . . . . . . . . . . . . . 15
2.4 Subsumption-based abduction problem . . . . . . . . . . . . . . . 15
2.5 Pareto-optimality . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.6 Directed hypergraph . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.7 Monoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.8 Directed hyperpath . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1 Relaxed abduction problem . . . . . . . . . . . . . . . . . . . . . 35
3.2 Irredundancy, dual irredundancy . . . . . . . . . . . . . . . . . . 35
3.3 Strict relaxed abduction problem . . . . . . . . . . . . . . . . . . 36
3.4 Entailment-based relaxed abduction problem . . . . . . . . . . . 45
3.5 T -entailment-based relaxed abduction problem . . . . . . . . . . 47
3.6 Normal form for relaxed abduction problems . . . . . . . . . . . 58
3.7 Induced hypergraph HRAP . . . . . . . . . . . . . . . . . . . . . 62
3.8 Rooted hyperpath weight . . . . . . . . . . . . . . . . . . . . . . 65
3.9 Safety of for pruning . . . . . . . . . . . . . . . . . . . . . . . 75
XVI Other Listings
List of Theorems
2.1 Axiom-based can simulate subsumption-based abd. . . . . . . . . 16
2.2 Subsumption-based can simulate concept abd. . . . . . . . . . . . 16
2.3 Cross product of (pre-)orders . . . . . . . . . . . . . . . . . . . . 21
2.4 Weakly Pareto-optimal elements . . . . . . . . . . . . . . . . . . 23
2.5 Strongly Pareto-optimal elements . . . . . . . . . . . . . . . . . . 23
3.1 Conservativeness of the relaxation . . . . . . . . . . . . . . . . . 36
3.2 Weak Pareto-optimality of RAP-solutions . . . . . . . . . . . . . 37
3.3 SolRAP size limit for strict RAP . . . . . . . . . . . . . . . . . . 38
3.4 Properties for inclusion-based orders . . . . . . . . . . . . . . . . 42
3.5 (Anti-)monotonicity of preorders . . . . . . . . . . . . . . . . . . 46
3.6 Properties for entailment-based orders . . . . . . . . . . . . . . . 46
3.7 T -strengthening for entailment-based orders . . . . . . . . . . . . 47
3.8 Properties for cardinality-based orders . . . . . . . . . . . . . . . 48
3.9 Relaxed abduction can simulate pinpointing . . . . . . . . . . . . 50
3.10 Correctness of RAPsolve-Generic . . . . . . . . . . . . . . . . . 52
3.11 Complexity of RAPsolve-Generic . . . . . . . . . . . . . . . . . 54
3.12 Correctness of RAPnorm . . . . . . . . . . . . . . . . . . . . . . . 60
3.13 Hyperpaths correspond to derivations . . . . . . . . . . . . . . . 64
3.14 Hyperpaths with minimum rooted weight are solutions . . . . . . 65
3.15 Termination and correctness of RAPsolve-EL+ . . . . . . . . . . . 67
3.16 Complexity of RAPsolve-EL+ . . . . . . . . . . . . . . . . . . . . 73
3.17 Pruning-safety of selected preorders . . . . . . . . . . . . . . . . 75
3.18 Effects of pruning . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.19 Complexity of RAPsolve-ELPlus with safe pruning . . . . . . . . 79
3.20 Correctness and complexity of RAPsolve-EL++ . . . . . . . . . . 87
Other Listings XVII
List of Algorithms
3.1 RAPsolve-Generic . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.2 RAPnorm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3 RAPsolve-EL+ . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.4 Procedure applyCR1 . . . . . . . . . . . . . . . . . . . . . . . . 69
3.10 Function join . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.11 Function meet . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.12 Function meet-closure . . . . . . . . . . . . . . . . . . . . . . . 72
3.13 Function remove-dominated . . . . . . . . . . . . . . . . . . . . 72
3.14 Function join-and-prune . . . . . . . . . . . . . . . . . . . . . . 75
3.15 Function join-and-prune’ . . . . . . . . . . . . . . . . . . . . . 78
3.16 RAPsolve-EL++ . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.17 Procedure applyCR1’ . . . . . . . . . . . . . . . . . . . . . . . 89
3.22 Procedure add-nogoods . . . . . . . . . . . . . . . . . . . . . . 91
3.23 Function join-and-prune” . . . . . . . . . . . . . . . . . . . . . 92
3.24 Function remove-dominated’ . . . . . . . . . . . . . . . . . . . 92
1 Introduction
1.1 Motivation
The Semantic Web vision in its purest form is to make the World Wide Web
machine-understandable. The Semantic Web community wants to realize
this vision by annotating all types of resources, e. g. text or hypertext
documents, images, and audio files with metadata that can be processed
automatically by computers. This information processing is facilitated by
technologies such as ontologies providing background information about an
ever growing number of domains, and automated reasoning employed to
evaluate information represented using such a formalisation of the domain.
Although the ultimate goal of having computers automatically search the
web and provide the interpreted results to the user has not been achieved yet,
current developments like the introduction of virtual personal agents into
smartphones show that the knowledge representation and reasoning (KR&R)
technologies underpinning the Semantic Web have reached an impressive
level of maturity.
Over the last years, KR&R technologies have found application in fields far
beyond this original vision, such as the analysis of multimedia data, diagnostic
services for complex industrial systems, or systems for supporting elderly
people in their home environment, to name only a few. Despite their apparent
diversity, all these applications can be understood as instances of a more
general information interpretation task: Given potentially large amounts
of observational data (often acquired from sensors including “intelligent
sensors” that provide preprocessed data) and a description of the laws
governing the specific domain, the goal is to determine a situation description
which explains these observations. Declarative descriptions of the domain
with formal semantics (often called semantic models or ontologies) are
conveniently used to express the underlying relationships between situations
and observations as well as additional background knowledge on the domain,
for instance object descriptions, structural information on machinery, or the
layout of an elderly person’s home. In this domain of model-based information
interpretation, the notion of a model is frequently used to designate both a

T. Hubauer, Relaxed Abduction,
DOI 10.1007/978-3-658-14407-4_1
2 1 Introduction
formal representation of a domain and an interpretation of information in

the sense of model-theoretic semantics. In this thesis, we adopt this general
approach, however, in places where confusion between the two meanings
might arise otherwise, we employ terms such as “knowledge base” or “domain
formalization” to designate the first interpretation.
The most prominent reasoning paradigm applied in KR&R and the Se-
mantic Web is deduction: based on a set of axioms, determine what additional
information is entailed. Deductive inference is well-studied and virtually all
available reasoning tools are based on deduction, mainly due to the fact that
deduction always yields correct conclusions: as long as the input is free of
contradictions, deduction will never lead to any invalid conclusions.1 On the
downside, deduction is only applicable in situations where all relevant input
information is known, or can at least be estimated. A significant proportion
of information interpretation tasks however do not comply with this pre-
requisite, reasons for this being both fundamental (e. g. non-observability of
internal processes) and practical (e. g. sensor failure). Abductive reasoning
provides a solution to this challenge by deriving hypothetical explanations
for set of dedicated axioms (e. g. representing observations), and allowing
a user to test these potential solutions against new information later on.
This capability of handling incomplete observational data makes abduction
a powerful tool for most information interpretation tasks. However, weak or
faulty domain formalisations still pose a significant challenge to abductive
information interpretation, particularly since they may make it impossible
to explain individual observations. Whereas some early approaches fail to
find any solution at all once a single observation cannot be accounted for,
more recent methods have found solutions to address this common issue –
typically by allowing to simply assume observations which cannot be ex-
plained otherwise. These approaches inherently incur a tradeoff between
the “cost” of assuming missing information and the “benefit” of explaining a
certain observation. Despite their obvious appeal, all these solutions either
return overly complex, suboptimal, or incomplete solutions.2 This obviously
poses a significant burden on the use of abductive reasoning in addressing
real-world information interpretation problems.
This thesis reports on research in bringing abductive reasoning to industrial
application. It proposes a novel, flexible method for logic-based abduction
1 Obviously, whether this entailed information indeed matches the intuitive expectations
of a user hinges on the fact that the formalisation correctly captures reality. It is
therefore no surprise that, depending on the complexity of the domain to be captured
and the degree of modelling detail, such formalisations tend to be very large.
2 We discuss this topic in more detail in Section 3.2.1.
1.2 Research Objectives 3
over imperfect domain formalisations that avoids the drawbacks outlined

before, and shows how it can be implemented efficiently. The applicability of
the proposed method is investigated in two industrial use cases on diagnostics
of machinery and rule bases respectively. The domain formalizations of both
use cases can be understood to instantiate a common meta model which
was developed in accordance with established industry norms on industrial
diagnostics.
1.2 Research Objectives

The thesis addresses the challenge of bringing logic engineering in general
and, more specifically, abductive inference into industrial information inter-
pretation applications. Our main research objective is therefore to develop a
method for model-based information interpretation that addresses both obser-
vational and domain model incompleteness, can be practically implemented,
and easily applied in a wide range of industrial use cases.
First and foremost, we aim at providing a formally founded framework for
information interpretation that guarantees both soundness and completeness
of the result. This framework must therefore meet the following requirements:
R1 (Expressiveness)
To allow for a compact representation of information and verifiability
of results, the framework must support modelling of and reasoning
over complex relational structures based on formal semantics.
R2 (Robustness to observational incompleteness)
The framework must support devising analysis results of a situation
even if the observational state is incomplete (meaning that certain
observations characterising a situation are unknown, e. g. due to sensor
failure).
R3 (Robustness to imperfect formalisations)
The framework must support devising analysis results of a situation
with respect to incomplete or faulty domain representations (e. g. due
to deficient expert knowledge on the domain under consideration).
R4 (Optimality)
To minimize distraction, the framework must not return suboptimal
solutions (with respect to user-definable optimality criteria).
4 1 Introduction
R5 (Coverage)
To maximize information gain, the framework must explore the full
spectrum of solutions in the area of tension between explanation
complexity and observation coverage.
Such a formal framework, however, defines characteristics of both the domain
representation and the solutions on a rather abstract level. To be of practical
use, it needs to be complemented with an algorithm for effective and efficient
generation of solutions. Requirements for this algorithm are:
R6 (Correctness)
The algorithm must be sound, complete, and computable in finite
time.
R7 (Efficiency)
The algorithm must allow for an efficient implementation. (Note that
in practice this does not necessarily require that the algorithm has
polynomial complexity. Numerous examples show that algorithms
with well-behaved exponential runtime can be applied very successfully
in practice.)
Taken together, these requirements characterise an information interpreta-
tion methodology that allows to make best use of (potentially incomplete)
information in a correct and efficient manner. Similar requirements can be
found e. g. in Dressler & Puppe (1999), stressing the relevance of the criteria
listed above. The knowledge-based nature makes the proposed solution
apt for a wide range of practical applications including, but not limited to,
diagnostic tasks in industry. We now introduce a simple diagnostics scenario
that will serve as a motivating example throughout this thesis. Next, the
subsequent section gives a short summary on how this work contributes to a
number of scientific fields.
Example 1.1 (A simple diagnostics scenario)
In our motivating example, we consider a simplified production system
equipped with a diagnostic unit. The system under consideration comprises at
least two independent subsystems, namely communications and transportation.
The former provides PROFINET connectivity to other devices in the factory,
the latter a mechanical gripper that can be used to move products from
the storage location onto the internal conveyor belt. The whole system is
governed by a central main control unit (or MCU, for short). Experts working
with the system have observed that a fluctuating power supply manifests
by intermittent outages of the main control unit while the communication
links remain functional and the mechanical gripper of the production system
is unaffected (the observations entailed by the diagnosis). Similarly, the
1.3 Scientific Contributions 5
maintenance manual of the system states that bugs in the MCU software
typically lead to intermittent MCU outages as well, accompanied by irregular
(i. e. unintended) derivations in the movement patterns of both gripper and
conveyor belt.
To minimize downtimes, it is essential to identify faults quickly and reliably
based on the available measurements from sensors attached to the system.
This task is to be accomplished by the diagnostic unit, based on automated
(abductive) reasoning.
1.3 Scientific Contributions

To address Requirements R1 to R7, this thesis contributes to the fields of
logic-based reasoning (with focus on description logics), knowledge repres-
entation, and industrial application of logic engineering. More specifically,
our contributions include the following:
• We identify the adequate treatment of imperfect domain formalisations
as an inherent flaw of standard formulations of logic-based abduction.
• We introduce relaxed abduction as a natural extension of abduction
that is capable of simultaneously addressing incompleteness of obser-
vations and domain representation in a sound and significantly more
flexible way than existing solutions do.
• We analyse the formal properties of this reasoning method, clarify its
relation to standard abduction, and investigate the effects of different
criteria for solution optimality.
• We show how relaxed abduction problems can be solved both for
general description logics and – using a more involved algorithm –
for the tractable description logic EL+ , and analyse the respective
computational complexity of the algorithms provided.
• We examine extensions to the EL+ algorithm, namely possible ex-
tension to more expressive modelling languages as well as efficient
support for the addition and retraction of axioms.
• We investigate how different types of information interpretation prob-
lems can be mapped formally into the relaxed abduction framework,
thereby strengthening the link between our novel formalism and real
industrial applications.
Summarising our contributions, this thesis makes a significant step in
closing the gap between theoretical results in the field of knowledge rep-
resentation and reasoning, and their practical applicability in the field of
6 1 Introduction
industrial information interpretation. By formally relaxing the well-studied

formalism of logic-based abduction, we significantly extend its applicability
in real-world applications without sacrificing its solid theoretical foundations.
Moreover, our studies on extensions and on a formal mapping to an ISO-
standardised diagnostic vocabulary help tighten the bound between theory
and successful application. Activities concerned with the dissemination
of the results developed in the course of this thesis are presented in the
subsequent section.
1.4 Dissemination Activities

This section summarises activities connected to the dissemination of the
results provided by this thesis. The presentation is clustered by types of
contributions.
Conference Papers
• Hubauer et al. (2011b) introduces an agent-based system for flexible
manufacturing control using a diagnostics unit based on relaxed ab-
duction. The results contributed mostly to the use case presentations
in Sections 4.3.1 and 4.3.2.
• Legat et al. (2011) complements Hubauer et al. (2011b) with in-
depth information on the formalisms used for automatic assessment
of remaining machine capabilities.
• Hubauer et al. (2012) presents a conceptual model for diagnostics
motivated by ISO standards, and a mapping between this formalization
and the proposed notion of relaxed abduction. It furthermore provides
evaluation results in the context of the turbine diagnostics use case.
The contents of this paper contributed to Sections 3.1, 4.2 and 4.3.1.
• Grimm et al. (2012) proposes an approach for bringing completion-
based reasoning to embedded devices such as programmable logic
controllers (PLCs), where computations must adhere to strict cycle
time restrictions. Though currently limited to classification of EL++
ontologies, the results gave valuable input to the implementation
of the RAbIT system presented in Section 4.1 and may serve as a
starting point for future work in the direction of embedded (relaxed)
abduction.
1.5 Outline 7
Workshop Papers
• Hubauer et al. (2010) focuses on standard logic-based abduction in the
context of description logics. It laid foundations for the hypergraph-
motivated analysis of abduction and, based on that, relaxed abduction.
• Hubauer et al. (2011a) motivates and introduces the framework of
relaxed abduction. It provides a first analysis of its properties as
well as the algorithm for solving relaxed abduction for EL+ and first
complexity results (which have been tightened meanwhile). This
publication mainly contributed to Sections 3.2 and 3.3.2.
Other Results
• The algorithm for solving relaxed abduction over EL++ knowledge
bases has been implemented in a Java component named RAbIT (an
acronym for Robust Abductive I nference T ool).
1.5 Outline
The remainder of this thesis is structured as follows: In Chapter 2 we intro-
duce basics on ontologies and description logics, abductive reasoning, order
theory, and hypergraphs. Chapter 3 steps right into the core contributions
of this thesis: After introducing a DL-based formalisation of diagnostics
(Section 3.1), we present the framework of relaxed abduction in general (Sec-
tion 3.2) as well as our algorithms for solving relaxed abduction problems in
the context of description logics (Section 3.3). Moreover, extensions to the
basic setting are considered in Section 3.4, and the chapter concludes with
a comparison to related approaches in Section 3.5. Chapter 4 presents the
RAbIT system, our implementation of the EL+ -based algorithm for solving
relaxed abduction, and analyses its performance (Sections 4.1 and 4.2).
Next, two case studies on applying relaxed abduction to problems from the
diagnostics domain are presented in Section 4.3. We conclude in Chapter 5
with a summary of our results and interesting open research questions.
2 Preliminaries
This chapter introduces topics that are relevant in the course of this thesis; the
presentations are hereby restricted to the aspects required in the remainder
of this work and not intended to be complete. Where needed to simplify
presentation later on, we also introduce new terminology and prove related
lemmas to simplify proofs later on. Section 2.1 gives a short overview on
ontologies and their relation to description logics as well as classical reasoning
tasks for DLs. Section 2.2 complements this account with an introduction
to abductive reasoning both from a general perspective and in the context
of description logics. Alternative definitions of abduction from literature are
introduced, and their relation to the definition given here is formally analysed.
As solving certain classes of relaxed abduction problems will be reduced to
finding optimal hyperpaths later on, a number of relevant order-theoretic
notions is introduced in Section 2.3; a basic account on hypergraphs, an
extension of standard graphs, can be found in Section 2.4.
2.1 Ontologies and Description Logics

Knowledge about the entities of a domain and their numerous interrelations
can conveniently be captured in an explicit and intelligible way by means of
an ontology. Representing domains by means of ontologies provides a deeper
understanding of the domain as compared to subsymbolic approaches or
other representations acquired by learning that often do not support the
representation of relational information. This makes it possible for domain
experts to gradually extend and adapt the formalisation to integrate their
knowledge. Nowadays, the method of choice for representing ontologies is
the Web Ontology Language (OWL) which is semantically founded on the
description logic (DL) formalism described in Nardi et al. (2003).
Description logics are a family of formal knowledge representation form-
alisms and can be understood as fragments of first-order logic restricted
to at most binary predicates. The main representational elements are con-
cepts (unary predicates), roles (binary predicates), and individuals (terms).
Description logics have model-theoretic semantics defined by means of an

T. Hubauer, Relaxed Abduction,
DOI 10.1007/978-3-658-14407-4_2
10 2 Preliminaries
interpretation function ·I that maps individuals to elements of the domain

Δ, concepts to sets of elements, and roles to sets of element pairs. The
entailment relation between a set X of DL axioms and a single DL axiom
x, denoted X |= x, is defined
as usual. For two sets of axioms X, Y , we let
X |= Y stand for X |= Y or equivalently: X |= y for all y ∈ Y . X =|Y
stands for Y |= X, and X ≡ Y is an abbreviation for X =|Y and X |= Y .
Later in this thesis, we need the intuition of an axiom set not containing
any redundant axioms. To capture this concept, we define the notion of a
logically independent set of axioms as follows:
Definition 2.1 (Logically independent axiom set)

Let X be a set of axioms with subsets S and S . Then X is logically
independent if and only if S |= S implies S ⊆ S .
In other words, an axiom set X is logically independent if no subset S ⊆ X
is redundant in the sense specified by Grimm & Wissmann (2011).
A concrete description logic is characterized by the constructors avail-
able for forming complex concepts from simple ones, and types of axioms
that can be stated. This choice typically constitutes a tradeoff between
expressiveness of the resulting language and computational complexity of
deciding satisfiability of a set of axioms called a knowledge base (KB) or
ontology. This task also known as consistency checking is one of the central
reasoning tasks in description logics, its decidability is a property shared
by all DLs. There is a number of other standard inference tasks which can
be solved by reduction to consistency checking for most description logics1 :
subsumption checking (determine whether every instance of concept C must
necessarily be an instance of concept D as well, denoted C D), classification
(determine all subsumption relations between concepts of the knowledge
base)2 , instance checking (determine whether individual i is an instance of a
concept C, denoted C(i) or i : C), and realization (for every individual of the
knowledge base determine the most specific concept names it is an instance
of).
1 Basically, reducibility for concept-oriented reasoning tasks rests on the capability of
expressing disjointness of concepts, whereas reducibility of individual-oriented tasks
also requires the existence of nominals, or enumerated concepts.
2 It should be pointed out that this definition coming from the community around light-
weight description logics differs from the standard definition used in most other parts
of the DL community, where classification determines only the most specific subsump-
tion relations between any pair of concepts. The difference can be understood as an
additional filter step on the full set of all subsumptions, retaining only the most spe-
cific ones. Therefore, classification according to the standard definition is typically
somewhat more expensive from a computational perspective.
2.1 Ontologies and Description Logics 11
In its second version OWL 2 (W3C OWL Working Group, 2009a), the
Web Ontology Language is largely based on the highly expressive description
logic SROIQ (Horrocks et al., 2006). SROIQ provides many powerful
features including enumerated classes {i 1 , . . . , i k }, full negation ¬C, various
qualified restrictions, role hierarchies, and inverse roles. This however
leads to high computational cost of the reasoning tasks introduced before.
Deciding concept satisfiability was shown to be N2ExpTime-complete in
Kazakov (2008). Since the doubly exponential complexity of SROIQ may be
prohibitive for certain applications, OWL 2 comes with a number of so-called
profiles which offer language variants with reduced expressiveness and better
computational properties. One of these profiles is OWL 2 EL, which has been
shown to be especially well suited for the representation of large taxonomic
structures such as the well-known SNOMED-CT3 and GALEN4 medical
ontologies. The EL profile is based on the description logic EL++ (Baader
et al., 2005a), for which consistency is decidable in PTime. This significant
reduction of complexity is made possible by the absence of a number of
constructors such as universal quantification, full negation and disjointness
axioms for concepts and roles, along with a number of other constructs that
could be used to express one of the previous. It has been shown by Baader
et al. that EL++ is maximal in the sense that none of these dropped features
can be included without sacrificing tractability. Despite these syntactic
restrictions, EL++ supports expressing information about individuals by
means of nominals (concepts comprised of exactly one named individual),
which allows to reduce reasoning over individuals to standard concept-level
reasoning. The same complexity bounds hold for the sublanguage EL+ which
has been proved sufficient for the representation of diagnostic ontologies in
various contexts (see e. g. Baader et al. (2007)). An EL+ ontology T is a
set of concept inclusion axioms of the form Turbine ∃ propelledBy . Gas
GasTurbine (which states that a gas turbine is a turbine propelled by gas)
and role inclusion axioms of the form directlyControls ◦ hasSubComponent
controls (stating that control is propagated over the partonomy of a system).
Table 2.1 summarizes the syntax and semantics of the concept constructors
and the axioms that can be represented in EL+ (where A is an atomic
concept, C, D are arbitrary concepts, and ri , r, s are roles). For details on
the description logic constructors we refer the reader to Nardi et al. (2003);
W3C OWL Working Group (2009b).
For making explicit that an axiom ax is expressed in the description logic
3 http://www.nlm.nih.gov/research/umls/Snomed/snomed_main.html
4 http://www.openclinical.org/prj_galen.html
12 2 Preliminaries
Table 2.1: EL+ syntax and semantics
Syntax Semantics
top concept ΔI
atomic concept A AI
concept conjunction C D CI ∩ DI
existential restriction ∃ r . C {x ∈ ΔI | ∃y ∈ ΔI :
(x, y) ∈ rI ∧ y ∈ CI }
concept inclusion axiom C D CI ⊆ DI
role inclusion axiom r1 ◦ · · · ◦ rn s r1I ◦ · · · ◦ rnI ⊆ sI
L we sometimes call ax an L-axiom. Analogously, a L-knowledge base is a

set of L-axioms. Given a description logic knowledge base T , we denote the
set of all role names occurring in T by NR , the set of all concept names by
NC , and define NC = NC ∪ { }.
2.2 Abductive Reasoning

This section provides an overview on abductive reasoning. We first give a
general introduction on the framework of (axiom-based) abduction and its
instantiations. Next, we relate this definition to a relatively new formulation
called concept abduction and show how the former subsumes the latter.
The section concludes with an analysis of the shortcomings of logic-based
abduction with respect to the processing of deficient domain representations,
one of the main challenges addressed in this thesis.
2.2.1 Standard Notions of Abduction

Abduction was introduced in philosophy of logic in the late 19th century by
Charles Sanders Pierce, who characterized it as the sole reasoning method
generating new information (Hartshorne & Weiss, 1931). It was rediscovered
as a method for determining diagnoses in the 1970s by Pople (Pople, 1973),
and has since then been employed as a method for interpreting incomplete
information in various applications such as text interpretation (Hobbs et al.,
1993), plan generation and analysis (Appelt & Pollack, 1992), and analysis
of sensor (Shanahan, 2005) or multimedia data (Möller & Neumann, 2008;
Peraldi et al., 2007). In the spirit of Pierce, abduction can be seen as
2.2 Abductive Reasoning 13
an inference scheme capable of providing possible explanations for some

observation. It is conveniently represented by the rule
φ⊃ω ω
φΘ
and interpreted as an inversion of the well-known generalised modus pon-
ens rule. This new derivation rule allows to derive φΘ as a hypothetical
explanation for the occurrence of ω , given that the presence of φ in some
sense justifies ω (where Θ is a unifier5 for ω and ω ). Such explanations
need not be unique, and while each single explanation is typically required
to fulfil some consistency requirement, different competing explanations may
well be contradictory to each other. Hypotheses can be falsified by adding
information (e. g. representing additional observations), making abduction
an inherently non-monotonic reasoning scheme.
Note that this general notion of abduction does not presuppose any
causality between φ and ω, as indicated by the notation φ ⊃ ω of the major
premise. On the contrary, various notions of how φ sanctions the presence
of ω give rise to different definitions of abduction as explicated in detail in
Paul (1993):
• Set-cover abduction implements the major hypothesis φ ⊃ ω by means
of an explicit mapping e : Φ → Ω called explanatory power that relates
sets of hypotheses to the sets of observations they explain. Determining
abductive solutions can then be reduced to finding minimum covers
of the observation set, based on the explanatory power. For realistic
examples however, the mapping e is prone to get huge and it is unlikely
that it can be defined correctly by hand, let alone updated on changes
in the domain.
• Logic-based abduction can be understood to replace the explicit map-
ping used in the set-cover approach by a knowledge base T , making the
full power of the formal language available for representing expressive
yet concise mappings. In addition to the domain formalisation, a set
A of abducible axioms is provided. A logic-based abduction problem
is solved by determining subsets of A which are consistent with the
axiom set T and explain the observation(s) given T . Hypothesis
construction for logic-based abduction is typically more involved than
in the set-cover variant since the mapping e is only known implicitly.
5 Giventerms ω1 , . . . , ωn , a substitution is a mapping from variables to terms which can
be understood as a generalised type of renaming, replacing (some of the) variables
with potentially complex terms comprising constants, variables, and functions. A
unifier Θ for ω1 , . . . , ωn is a special substitution such that ω1 Θ = . . . = ωn Θ.
14 2 Preliminaries
• Knowledge-level abduction generalizes logic-based abduction with a

conceptual model of belief. In a nutshell, φ explains ω w. r. t. an
epistemic state if (in this state) it is believed that φ → ω, and the
negation of φ is not believed. This approach is claimed to be relevant
e. g. for situated agents where the epistemic state might change
regularly due to external factors.
Obviously, set-cover abduction will typically be too weak to meet Require-
ment R1. As representation of and reasoning over belief is not relevant for
the application scenarios at hand, we focus on logic-based reasoning over L-
knowledge bases (for decidable logics L), i. e. we understand abduction as a
non-standard inference task over formal, logic-based domain representations.
In this context, an abduction problem can be defined as follows (c.f. Eiter &
Gottlob, 1995):
Definition 2.2 (Abduction problem)

An abduction problem is a 3-tuple AP = (T , A, O), where
• T is a set of axioms that formalise the domain,
• A is a set of abducible axioms representing assumptions, and
• O is a set of axioms representing observations.
A solution to AP is a set A ⊆ A such that T ∪A is consistent and T ∪A |= O.
Naturally extending the standard notation, we write A |= AP to state that A
solves AP. The solution set for AP is defined by SolAP := {A | A |= AP}.
It should be pointed out that abduction does not at all require that every
conclusion of T ∪ A has been observed (i. e. is an element of O). This
separates abduction from mere backward-chaining rule-based deduction,
and enables it to flexibly address situations where the observation set is
incomplete.6
Due to the fact that an abduction problem does in general not have
one unique solution A but a collection of alternative answers A1 , . . . , Ak ,
one typically selects optimal solutions by means of a (not necessarily total)
preference order , with Ai Aj expressing that Ai is at least as good as
Aj (see Section 2.3 for an introduction to orders). Abductive inference can
then conveniently be formulated as an optimization problem as shown next:
6 Backward-chaining rule-based reasoning can obviously emulate this feature by breaking

down rules with multiple conjuncts in the precedent. However, this typically leads to
a significant explosion of the number of rules to consider.
Definition 2.3 (Preferential abduction problem)

A preferential abduction problem is a 4-tuple PrefAP = (T , A, O, A ),
where
• A is a set of abducible axioms representing assumptions,
• O is a set of axioms representing observations, and
• A ⊆ P(A) × P(A) is a partial order over sets of assumptions.
A pre-solution to PrefAP is a set A ⊆ A such that T ∪ A is consistent and
T ∪ A |= O. A solution is an A -minimal element of the set of pre-solutions.
We write A |= PrefAP to state that A solves PrefAP. The solution set for
PrefAP is defined by SolPrefAP := {A | A |= PrefAP}.
Typical orders over sets include subset minimality (Ai s Aj if and only if
Ai ⊆ Aj ), minimum cardinality (Ai c Aj if and only if |Ai | ≤ |Aj |), and
weight-based approaches defined by a function w that assigns numerical
weights to subsets of A (i. e. Ai w Aj if and only if w(Ai ) ≤ w(Aj )).
2.2.2 Concept-based Notions of Abduction

A different formulation of abduction due to Bienvenu (2008) employs atomic
concepts for representing the abducibles and the observation. Note that we
use the term subsumption-based abduction for this approach instead of the
original name concept-based abduction, in order to clearly distinguish it from
concept abduction introduced later. For similar reasons, we will from now
on use the term axiom-based abduction for the standard definition (Defini-
tion 2.2). After introducing subsumption-based abduction in Definition 2.4,
we show in Lemma 2.1 that this notion is less general than axiom abduction.
Definition 2.4 (Subsumption-based abduction problem)

A subsumption-based abduction problem is a 3-tuple SAP = (T , H, O),
where
• H is a set of atomic concepts (each representing a possible assumption),
and
• O is an atomic concept representing the observation.
A solution to SAP is a set H ⊆ H such that Ci ∈H Ci is consistent w. r. t.

T and T |= Ci ∈H Ci O.

16 2 Preliminaries
Lemma 2.1 (Axiom-based can simulate subsumption-based abd.)

A subsumption-based abduction problem SAP = (T , H, O) can be solved by
determining solutions to the corresponding axiom-based abduction problem
AP = (T , A, O) defined by:
• A = {C∗ Ci | Ci ∈ H},
• O = {C∗ O}, and
• C∗ is a new concept name, C∗ ∈ NC
More concretely, A = {C∗ Ci } solves AP if H = {Ci | C∗ Ci ∈ A}
solves SAP.
Proof. Let H solve SAP, i. e. T |= Ci ∈H Ci O. Then T ≡ T ∪

{ Ci ∈H Ci O}, and obviously T ∪ { Ci ∈H Ci O} ∪ {C∗ Ci | Ci ∈

H} |= C∗ O. Thus T ∪ A |= O, A therefore solves AP.
In Colucci et al. (2003) the similar term concept abduction is used for a
slightly different task: Given a L-knowledge base T and two L-concepts
C and D that are satisfiable in T , determine a L-concept H such that
T |= C H ⊥ and T |= C H D. This problem can be reduced to
subsumption-based abduction as shown in the following lemma:
Lemma 2.2 (Subsumption-based can simulate concept abd.)
Let (T , C, D) be a concept abduction problem over language L as defined
by Colucci et al. (2003). Then, all solutions H to (T , C, D) can be determined
using subsumption-based abduction as defined in Definition 2.4.
Proof. Define a subsumption-based abduction problem SAP for (T , C, D)
by SAP(T ,C,D) = (T , {C} ∪ L∗ (d), D), where L∗ (d) denotes the set of all
semantically distinct L-concepts of role depth ≤ d, with d defined as the
maximum role depth occurring in C, D, or any axiom of T . Further, let
SolSAP(T ,C,D) be the solution set for the induced subsumption-based abduc-
tion problem, and SolSAP(T ,C,D) |C its restrictions to subsets of H that do
contain C.
Assume that H solves SAP(T ,C,D) , and H ∈ SolSAP(T ,C,D) |C . Then H ⊆ H,
Ci ∈H Ci is consistent w. r. t. T (i. e. T |= Ci ∈H Ci ⊥), and T |=
Ci ∈H Ci D. Now define C H = Ci ∈H Ci . As C ∈ H by construction, it
is also true that T |= C CH ⊥ and T |= C CH D, i. e. H solves the
original concept abduction problem (T , C, D).
The notion of abduction introduced in Definition 2.2 is therefore non-

trivial and practically relevant, making it a good basis for our extension
presented in Chapter 3.
2.2.3 A Critique of Logic-Based Abduction

A joint property of all formalizations of abduction introduced so far (inde-
pendently of the concrete logic L used) is their requirement that a solution
must completely explain the observation(s): There is no way of having a
solution “explain the largest part of O” or “make C almost subsume D”.
Typical application fields, such as media analysis (Castano et al., 2009) and
industrial diagnostics (Hubauer et al., 2011b), are however characterized by
an abundance of low-level observations due to a large number of sensors,
whereas the knowledge base formalising the domain is often rough or incom-
plete since it must be created manually. As the following example shows,
existing definitions of abduction are not sufficient in this context:
Example 2.1 (Diagnostics over incomplete domain formalisations)
The simple production system introduced before consists of a main control
unit (MCU), a mechanical gripper and a conveyor which are both part of the
transportation subsystem, and a PROFINET communication link to other
systems of the factory. The system and its structure can be expressed using
EL+ axioms as follows:
MCU ∃ partOf . System
Communications ∃ subsystemOf . System
Transportation ∃ subsystemOf . System
PROFINET ∃ belongsTo . Communications
Gripper ∃ belongsTo . Transportation
Conveyor ∃ belongsTo . Transportation
belongsTo ◦ subsystemOf partOf
The diagnostic expert knowledge of the system laid out before informally in
Example 1.1 can then be formalized with the following set T of axioms:
MCU ∃ partOf . (∃ operatesIn . PowerSupplyFluctuations)
∃ shows . IntermittentOutages
PROFINET ∃ partOf . (∃ operatesIn . PowerSupplyFluctuations)
∃ shows . SendingReceivingOK
Gripper ∃ partOf . (∃ operatesIn . PowerSupplyFluctuations)
∃ shows . FullyFunctional
MCU ∃ partOf . (∃ operatesIn . ControlSWMalfunction)
∃ shows . IntermittentOutages
Another random document with
no related content on Scribd:
“And,” whispered Cicely, “we can feel that it is only we two who
suffer. My father has been spared all this.”
“Yes,” said Mrs. Methvyn, “we may be thankful for that.”
“And after a while,” continued Cicely, “you and I will go away
together to some new place where there be nothing to recall all this,
and we shall be very happy and peaceful in our own way, mother
dear, after all, shall we not?”
Mrs. Methvyn tried to answer cheerfully, but she could not
manage it. She only shook her head sorrowfully, while the tears ran
down her thin cheeks. Cicely kissed them away.
“I don’t know, dear,” the mother whispered. “I would not feel it so if
I could look forward to being able to do anything to make you happy
again. But I am getting old, my darling, trouble ages one, and I feel
as if half my life had gone with your father.”
“But after a while you will not feel it so bitterly,” persisted Cicely,
and Mrs. Methvyn tried to believe it would be so.
As far as was possible Cicely spared her mother all the painful
details of the utter change in their prospects. It was arranged that
Geneviève should return home to Hivèritz at the same time that the
Methvyns left the Abbey; and though before then it became
necessary to tell Mr. Fawcett’s parents of his engagement to Miss
Casalis, the fact was not made public. And fortunately for herself,
Geneviève’s spirits continued in a subdued state during the short
remainder of her stay.
“She looks so frightened and miserable, mother,” said Cicely one
day when Geneviève had rushed out of the room suddenly, to avoid
meeting her aunt, who came in unexpectedly. “Can you not forgive
her?”
“I have forgiven her,” said Mrs. Methvyn coldly. “I promised you I
would. I confess I have not yet come to pitying her, as you seem to
do, Cicely. I know her now better than you do. More than half of that
misery is affectation. She will not have a thought of self-reproach
when it comes to buying her trousseau and being congratulated. Oh!
I do wish that you had let my first letter go, the one in which I told
Caroline the truth.”
“But you did tell her the truth in the one that went,” said Cicely.
“You told her that my engagement had been broken off entirely by
my own wish, and that Mr. Fawcett had fallen in love with Geneviève,
and that his parents approved of it.”
“That was not the whole truth,” said Mrs. Methvyn. “Do not take
up my words in that way, Cicely.”
“Well, mother, there would have been no good in telling more,”
replied Cicely gravely. “What Geneviève may choose to tell her
mother herself is a different matter, and does not concern us. In the
same way I very much prefer that Trevor’s parents should be told
nothing by us, though I fear poor Sir Thomas suspects something.
How kind he has been!”
She sighed as she recalled her old friend’s endeavours to shake
her determination. Lady Frederica had cried over it till the thought of
the admiration that would certainly attend the débút in fashionable
society of so lovely a daughter-in-law as Geneviève, had suggested
itself as consolation. But long before there was any talk of his son’s
“new love,” Sir Thomas had come over to see Cicely in hopes of
getting to the bottom of the mysterious misunderstanding between
her and Trevor. He had gone away sorrowful, convinced at last that
the girl’s decision was unalterable, but none the more reconciled to it
on that account.
“I trust, my dear, that, as you assure me, false pride has nothing
to say to it,” were the parting words of the honest-hearted old man.
“Trevor told me it was no use my coming, but I—well, I fancied old
heads were cooler than young ones sometimes, even in a case of
this kind. I can’t blame you, Cicely—I think too highly of you for that,
even if my boy had not assured me most solemnly that what you
have done has only doubled his respect and admiration for you. If I
thought he was to blame” a hard look came over Sir Thomas’s
comfortable face.
“Don’t think any one is to blame,” Cicely entreated, “in the end we
may all come to see that it has been for the best.”
“And you knew about my thinking of buying Greystone?” he said
regretfully.
“Yes,” she replied. “It was a most kind thought. But I have
understood since that it would not be considered a profitable
purchase for you—you don’t want more land about here?”
“No, I don’t care about it. And I don’t care about another house
either. I have Barnstay up in the north, you know—the lease is nearly
run out; in case Trevor ever marries, he may live there if he likes. No,
I don’t want Greystone; but if you would still like the idea of my
having it, if there were the ghost of a chance of things ever coming
straight again between you and Trevor—”
“There is not the ghost of a chance, dear Sir Thomas,” replied
Cicely. “Do not let any thought of us influence you in the matter.
Greystone will be dead to me from the day we leave it.”
“That means you will never come back eh?” said Sir Thomas.
Cicely did not contradict him. “Ah! well, then, I think I’ll give up the
idea. It was different when I thought of keeping it together for poor
Philip’s grandchildren.”
He kissed Cicely when he left her, and there were tears in his
eyes, but the subjects they had discussed were never alluded to
again. Some weeks later when his consent to Trevor’s marriage was
asked; he gave it without difficulty. But he thought his own thoughts,
nevertheless; and it came to be generally noticed that the old
gentleman never “favoured” his pretty daughter-in-law as much as
might have been expected. “Not like it would have been with our
Miss Cicely,” the people about used to say.
But time went on. Greystone Abbey was sold to strangers, and
the desolate widow and daughter of its last owner left it for ever.
CHAPTER VI.
A NEW TERROR.
“It is not for what he would be to me now,

If he still were here, that I mourn him so:
It is for the thought of a broken vow,
And for what he was to me long ago.”
Lord Houghton.
LAST days are generally sad, sometimes terrible things. Sad enough
were these last days at Greystone: Cicely felt thankful when they
were over, though once gone, she would have given years of life to
recall them. It was a relief to both her mother and herself when
Geneviève left them, and they were free to take farewell of the
places they had loved so dearly, without the painful associations of
her presence, gentle and subdued as she remained to the end. It
was a relief to them, and they were conscious too that it could not
but be a relief to her.
“I shall tell my mother all,” she had said to Cicely, and Cicely was
glad to hear of her intention.
“You are right to tell her, Geneviève,” she said, “though no one
else would ever have done so.”
But Geneviève’s confession was deferred. She did not, after all,
return to Hivèritz as proposed, for there came news from Madame
Casalis of a fever having broken out in the household, which
threatened to be of a serious kind and without doubt infectious; one
of the boys and little Eudoxie were already prostrated by it; the latter,
it was feared, likely to have it badly. Under these circumstances,
therefore, it was proposed by Geneviève’s parents that she should
only go as far as Paris and stay there under the care of an old school
friend of Madame Casalis, who was by no means averse to receiving
as a visitor the ‘promise’ of a wealthy young ‘milord.’ Geneviève
shed tears over her mother’s letter, expressed herself désolée at the
thought of the anxiety in her family, but ended by deciding that her
persisting in returning home to the modest little house in the Rue de
la Croix blanche, would only add to her mother’s trouble and
distress.
“It is best I should go to Madame Du plessis,” she said resignedly.
“As mamma wishes it, it is best I should not ask still to go home.”
And her satisfaction with her own decision was not a little
increased when, the night before she was to leave, her aunt gave
her, to expend upon her trousseau, a sum of money which was now
to Mrs. Methvyn by no means the trifle it would have been a few
weeks before.
“She is my relation, however she has behaved,” said Cicely’s
mother. “It would never do for her to enter the Fawcetts’ family
without a proper outfit.”
“And I can get everything in Paris,” thought Geneviève delightedly.
And her expressions of gratitude were so evidently sincere that even
Mrs. Methvyn’s heart was a little softened to her, and she bade her
good-bye with more kindliness of manner than might have been
expected.
It ended in Geneviève’s remaining in Paris till her marriage—her
father and mother joining her there only a few days before it took
place. How much or how little, therefore, of the true history of her
daughter’s love affairs was confided to Madame Casalis, Cicely did
not till long afterwards know, though from the tone of her letters to
Mrs. Methvyn it was evident that “poor Caroline’s” satisfaction in
Geneviève’s brilliant prospects was not unalloyed.
“I would not conceal from you, my dear and kind Helen,” she
wrote, “that I am glad for my child to be well settled in life. But my
grief and sorrow for you prevent my being able to rejoice as I might
otherwise have done. And though you tell me, and it is no doubt the
case that the breaking off of your Cicely’s engagement was by her
own desire, I cannot rid myself of the feeling that it is unnatural and
strange that Geneviève should so quickly have succeeded her in Mr.
Fawcett’s regard. I trust neither he nor my daughter may regret what
seems to have been done hastily. She tells me she will explain all
when we meet; she says your goodness and that of the dear Cicely
is beyond words. I trust, when I understand more, I may not learn
that my child has not been deserving of it.”
“She is so simple and unworldly!” said Mrs. Methvyn. “Poor
Caroline—how can her daughter have become so different!”
But after reading this letter, she did not again repeat her wish that
Cicely had allowed her to tell Geneviève’s mother the whole truth.
When they first left the Abbey, it was the intention of Mrs. Methvyn
and her daughter at once to make a new home for themselves at
Leobury, a quaintly pretty, quiet little town, far away from Greystone
and its associations. It was at Leobury that Mrs. Methvyn’s former
widowhood had been spent, and it was there too that after a
separation of ten years she had again met her first love—Philip
Methvyn. Leobury was endeared to her for this reason, and Cicely
was only too glad to find her mother able to express interest on the
subject of their home, not to consent readily to what she proposed.
But their plans were not at once carried out. There was no suitable
house to be had at Leobury at the time they left Greystone. So for
nearly a year after Colonel Methvyn’s death, his widow and daughter
were without a settled home.
Summer was again on the wane—it had seemed a short summer
to Cicely compared with that of the previous year—when they at last
found themselves in possession of a pretty little house in the place
where, more than a quarter of a century before, young Mrs. Bruce
had come to live with her two years old Amiel. She had been poor
then; poorer far than now, but in those old days poverty had seemed
a smaller matter. She was young, and she had then never known
what it was to be rich; the husband she had lost had caused her far
sorer griefs than that of his death—she had suffered enough in her
short married life to realise even loneliness as a relief. Now, how
different everything was! For twenty-two years she had been happier
than it falls to the lot of many women to be; hardly a wish of hers had
remained ungratified; till Colonel Methvyn’s accident she had not
known a care—when suddenly the whole had collapsed—for the
second time she found herself a widow, and in comparatively
speaking straitened circumstances, her companion this time a
daughter even dearer to her than little Amiel—a daughter who,
young as she was, had known sorrow and disappointment neither
slight nor passing, troubles enough to have soured a less healthy
nature, which her mother was powerless to soften or heal. It is hardly
to be wondered at that, weakened in strength and nerve, Mrs.
Methvyn sometimes found it difficult to look on the bright side of
things.
But what she could not do, time and nature had not neglected.
Already Cicely was recovering from the crushing effects of the blow
that had fallen upon her so unexpectedly. She could look back now
without flinching upon all that had happened, could feel that she was
still young and strong and hopeful, and that the light had not all died
out of her life. To her their loss of riches had been by no means an
unmitigated evil; for though they were not so reduced as to come
into unlovely proximity with poverty, the change in their
circumstances had necessitated the exertion of energy and
forethought in their arrangements, which had been for Cicely a
healthy and invigorating discipline. She felt that she was of use; that,
without her, life would have been a terrible blank to her enfeebled
mother; and the consciousness gave her, as nothing else could have
done, strength and cheerfulness. One thought only she could not
face—what would life be to her now, how would it be possible to
endure it were her one interest removed, her mother taken from her?
“I am risking my all in a frail bark,” she would sometimes say to
herself with a shudder, “but surely if mamma were to die I should die
too. No one could live in such utter desolation.” Yet even as she said
so, a misgiving would suggest itself that such things had been and
might be again; that life, the awful yet priceless gift, though bestowed
unasked, is not therefore so easily to be laid down as at times we
would fain imagine. The “sharp malady” must run its course; the
“appointed bounds” cannot be passed; the “few days and full of
trouble” are to some lonely men and women lengthened into the
many, before there comes the longed-for order of release, the
permission to depart for the unseen land which, wherever and
whatever it may be, is to them more real than any other, since thither
have journeyed before them all whose presence made this earth a
home—the “faces loved and lost awhile;”—yes, there are some
strangely solitary beings in this crowded world of ours.
The first two or three Sundays at Leobury, Cicely was not able to
go to church. Her mother was not very well and shrank from being
left alone. But, at last, by the end of September, there came a
Sunday when there was nothing to prevent Cicely’s at tending the
morning service in the fine old church, but a stone’s throw from her
mother’s house.
It was a beautiful day, soft and balmy; yet already, through the
trees and in the breezes, there came the first far-off notes of the
dying year’s lament—the subdued hush of autumn was perceptible.
“The swallows have gone, I feel sure,” thought Cicely, as she
walked slowly up the quiet village street (for after all Leobury was
hardly deserving of the name of town). “I can always feel that they
are gone.”
She gazed up into the sky, the blue was mellowed and yet paled
by the golden haze of harvest-time. “The summer has gone again,”
she murmured as she passed into the church by an old porch-way
with stone seats at each side, which reminded her of the entrance to
Greystone Abbey—“the old porch at home”—and it was in a sort of
dream that she made her way to a seat in a quiet corner. She hardly
noticed when the service began; she stood up mechanically with
those about her, but her thoughts were far away. Suddenly, without
knowing what had suggested it, she found herself thinking of that
Sunday morning, more than a year ago now, in Lingthurst church—
the first Sunday after Geneviève’s arrival. She remembered how the
sunshine bad streamed into the ugly little building, brightening into
brilliant pink the old woman’s brick dust cloaks, making also more
conspicuous the mildew stains and patches on the plastered walls.
She recalled her feelings of commiseration for the new clergyman on
this his first introduction to his church. What had put all this into her
mind just now? There was nothing in the place to recall Lingthurst—
Leobury church was as picturesque and impressive as Lingthurst
was commonplace; the one was perfect of its kind, the other
glaringly bare and unattractive—the familiar words of the service had
been listened to by Cicely in many churches during the last few
months without recalling any special association. Suddenly the riddle
was solved Two clergymen were officiating; the one, an old white-
headed man with a feeble quavering voice, which contrasted
curiously with the firm clear tones of his assistant priest. From where
Cicely sat, the younger man was not visible, but by moving a little
she obtained a view of him, and understood the trick which the
“quaint witch” had been playing her—she saw before her the grave,
boyish face of Mr. Hayle!
How had he come there? How strange it seemed! A little shiver
passed through her as she recalled the last time she had seen him—
it had been on the night of the Lingthurst ball, the night so full of
misery for her. She had never seen him since then, for though he
had called at the Abbey after her father’s death, she had shrunk from
meeting him again, though far from ungrateful for his kindness and
consideration. And now here he was at Leobury!
“I shall not mind seeing him again now,” thought Cicely, “I think I
am rather glad he is here. Mother liked him. I don’t think she will
have any painful feeling to him, poor little man!”
But the rest of the service passed like a dream. Cicely’s thoughts
were away in the past—wandering in the land of long ago.” She
recalled her happy childhood, her girl hood so full of love and
promise—the sunny days when trouble and sorrow seemed such
remote, all but impossible, possibilities! How Trevor seemed
associated with it all—there was not a walk she had ever taken, not
a summer ramble or winter skating expedition in which his figure did
not seem prominent.
“I wish he had been my brother, really,” sighed Cicely, “then no
one and nothing could ever have separated us!”
It was a torn out page—a page which it would ever be painful to
miss. But she was beginning to realise that the book held others—
others which hereafter she would not shrink from looking back to and
lingering over with loving tenderness of remembrance.
When the service was over, Cicely walked slowly homewards.
She was near her mother’s house, when the sound of her own name
made her look round. Mr. Hayle was behind her.
“Miss Methvyn,” he exclaimed eagerly, his face flushed with
pleasure and the quick rate at which he had been walking, “I thought
it was you. I was sure I could not be mistaken.”
“Did you see me in church?” said Cicely, shaking hands with him
as she spoke. She was pleased to see him, but again, at this first
moment of meeting, there rushed over her the remembrance of the
last time she had spoken to him, and unconsciously a slight
constraint showed itself in her manner.
“Yes,” said Mr. Hayle. “But I was on the look-out for you. I heard
yesterday evening quite accidentally that a lady of your name had
taken a house here, and I began to hope it might be you. I trust Mrs.
Methvyn is well. I—I hope I may come to see her?”
His tone had sobered down to its usual slightly formal gravity, and
his last words sounded somewhat ill-assured and hesitating. Cicely
felt vexed with herself for chilling him.
“I am sure mamma will be pleased to see you,” she said, “when I
tell her of your being here. I was so surprised to see you,” she
continued, “we did not know you had left Lingthurst.” She uttered the
word firmly and distinctly. Mr. Hayle looked more guilty than she.
“I left some months ago,” he said. “I am only here temporarily
however—staying with my uncle who is the rector of Leobury. He is
old and infirm, and I do what I can to help him.”
“Then have you no plans for the future? Why did you leave
Lingthurst? Was it not very sudden?” questioned Cicely.
“I got discouraged there,” he replied. “No one—after you left—
took any interest in things. There was very little I, alone and
unassisted, could do, and that little a less strong man—perhaps I
may say a less energetic man—could do quite as well or better. So I
gave it up. I have plans, but—”
“Did my cousin—Mr. Fawcett, I mean—did he take no interest in
things?” interrupted Cicely. She looked up in Mr. Hayle’s face without
flinching. He could see that she wanted an answer.
“No,” he said hesitatingly. “He is not much there now, and—and
when they are there Lingthurst is generally full of company. Mrs.
Fawcett is fond of amusement.”
“Yes,” said Cicely, “she is young—it is natural she should be. But
Trevor used to speak so much, after you came, about doing
something really to improve Notcotts and all that neglected part; I am
disappointed.”
“He may do so in time. He has had plenty of other things to think
of this last year,” said Mr. Hayle. Then he grew scarlet with fear that
Miss Methvyn should think he was alluding to the past. But
fortunately by this time they had reached Cicely’s home, and it was
easy for her to affect not to notice his discomposure.
“Will you come in?” she said, with her hand on the gate. “I think
mamma will be downstairs—it is half-past twelve—she is generally
dressed by this time.
“Is Mrs. Methvyn not as well as she used to be, then?” inquired
Mr. Hayle, recalling the early hours and active habits of Greystone.
“Oh! dear, yes; she is very well—perfectly well,” said Cicely
quickly. “She says she has grown lazy, that is all. Won’t you come in
and see her for yourself?”
“Not just now, thank you,” he replied. “If I may call to-morrow, I
should like very much to see Mrs. Methvyn.”
“Come to-morrow then, by all means,” said Cicely brightly. She
smiled as she spoke—she was so anxious to convince him of her
cheerfulness and well-being.
“Thank you,” he said simply, with something in his expression
which she did not understand. Then he shook hands again and went
away.
“What energy and powers of endurance she has,” he said to
himself. “Such a woman has it in her to do great things.”
Mrs. Methvyn was not in the drawing room when Cicely went in.
The girl ran upstairs.
“Mamma,” she called out, tapping at her mother’s door.
“Come in,” replied Mrs. Methvyn’s voice, and Cicely entered. Her
mother was dressed, sitting in an arm-chair near the fire.
“You are very lazy this morning, mother,” she said laughingly. “I
was very nearly bringing an old friend in to see you. Whom do you
think I met at church?”
“An old friend,” repeated Mrs. Methvyn. “Was it could it have been
Mr. Guildford?”
She spoke so eagerly that Cicely looked at her in surprise; and
now for the first time she observed that her mother was exceedingly
pale.
“Mother dear, I am afraid I have startled you,” she said penitently.
“No, it wasn’t Mr. Guildford. It was Mr. Hayle. Shall you dislike his
coming to see you to-morrow.”
“Oh! no. I shall be very glad to see him,” replied Mrs. Methvyn.
But the interest had died out of her voice. She leaned back in her
chair as if exhausted. “Will you see if Parker has come in, Cicely?”
she said. “I sent her out on a message a few minutes ago. I shall
come downstairs in a little while, and then you can tell me about Mr.
Hayle. I did not know he had left—Lingthurst.”
By her the word was pronounced with an evident effort. Just then
Parker came in. “I think this is the same as the last you had, ma’am,”
she said, “but it is not often different chemists prepare things quite
alike.”
She had a small phial in her hand, she did not observe Cicely
standing by.
“What have you been getting mamma medicine for, Parker?” she
said. “She’s not ill.”
Parker started. “It is only the same tonic that Dr. Farmer gave me
long ago,” said Mrs. Methvyn. “Go and take off your things, dear. I
shall be down directly.”
Cicely was not satisfied, but she left the room. Later in the
afternoon when she and her mother were alone together, she
recurred to what had been said. “Mamma,” she began, “what made
you think of Mr. Guildford to-day when I told you I had met an old
friend?”
“I think I should have been glad if it had been he,” said Mrs.
Methvyn. “I had great confidence in him. I think him very clever.”
“But what about it? You are not ill; you don’t want a doctor,” said
Cicely.
“I don’t know, dear. I don’t think I am very well,” replied Mrs.
Methvyn tremulously. “Long ago—years ago I used to have now and
then a sudden sort of attack, a kind of spasm, which some doctors
thought had to do with my heart. Then for some time these attacks
almost ceased, and I thought it must have been a mistake—but
lately they have returned much more frequently and violently than
before. And—Cicely dear—when I went up to town with Parker that
day from Brighton—I would not let you come, you know—it was to
see Dr.——, the great authority on that class of disease.”
“And what did he say?”
“He said that there was nothing to be done. It was true what I had
been told. He said that—that I might not get worse for a long time,
but that—it is not certain that I shall not.”
“Mamma,” cried Cicely gazing up in her mother’s face with wildly
agonised eyes, “mamma, you are going to die and leave me. That is
what you mean. Why didn’t you tell me before? oh! why didn’t you?”
“My child, my darling, I could not,” said her mother, the tears
coursing down her thin pale cheeks. “How could I break your heart
again? And it is only quite lately that I have begun to be afraid about
myself—only quite lately. It may pass off again—I may be with you
for many years yet.”
“No,” said Cicely, “no. I feel it coming. Mother, oh! mother, what
shall I do? How can God be so cruel? May not I die too? Oh,
mamma, mamma!”
She burst into an anguish of wild weeping, and for some moments
Mrs. Methvyn did not check her. Then at last she whispered, “Cicely
dearest, will you not try to be calm? For my sake—I cannot bear to
see you go.”
The words recalled Cicely to herself. Yes, she must be calm. The
pain of witnessing such stormy grief would assuredly weaken her
mother’s already feeble hold on life. With a violent effort she checked
her sobs, and fought for self-control. Then she listened to all her
mother had to say; listened, and pretended to believe that her fears
had exaggerated the danger; but in her heart of hearts she knew the
truth—the last drop of bitter sorrow had been poured into the cup of
which, for one so young, she had already drunk deeply.
When Mr. Hayle called the next day, he was shown into the
drawing-room and there found Mrs. Methvyn alone. He stayed with
her a considerable time, but he did not see Cicely at all. When he
came away from the house, his face looked sad and careworn, and
he sighed deeply: “Poor things, poor things!” he murmured to
himself. Yet he was well used to sorrow and suffering! “I will call
again in a day or two if you will allow me,” he said to Mrs. Methvyn at
parting. And she thanked him warmly and begged him to come.
“Cicely will be better again before long,” she said, “and she will like
to see you. It is the first shock that has, as it were, overwhelmed
her,” she added, with a sort of gentle apology of manner that touched
Mr. Hayle greatly. “I will come again soon,” he repeated.
And the next time he came, he found Cicely in the drawing-room
with her mother. They were sitting quietly working and talking, these
two all but broken-hearted women, as if no terrible tragedy hung over
them, as if they found life the easy, even, pleasant thing it looks to
some, till their time of trouble comes too, as in due course it must.
At first they talked about commonplace things—the weather and
Leobury and some of the people they had come to know there. Then
Mr. Hayle told the two ladies something of his own plans; how he
was likely to get a living in the east end of London, where there
would certainly be no lack of work for mind and body. At another time
Cicely would have listened with the greatest interest—even to-day
her sorrowful absorption of thought could not altogether enchain her.
“And you will throw yourself altogether utterly—into your work, I
suppose,” she said to Mr. Hayle. “I wish I were you. I suppose with
so intense an interest, no personal sorrow would be unendurable. If
you were without a friend on earth, you could still find happiness in a
life so spent.”
“I hope so—I believe so,” replied he, his eyes sparkling with
enthusiasm. “Or if not happiness, something better—better at least
than what we commonly understand by happiness—blessedness.”
“It is the old martyr spirit in another form, I suppose,” murmured
Cicely in a low voice. “I wish I could catch it.”
There fell a little pause. Then with a little hesitation the girl turned
to the young clergyman again.
“Mr. Hayle,” she said, “you remember Mr. Guildford, the
Sothernbay doctor who was so good to my father. You liked him, did
you not? I wonder if by any chance you have kept up
correspondence with him. I should so like to know where he is.”
“I can tell you where he is,” said Mr. Hayle. “I did hear from him
once after he left Sothernshire. Yes I liked him very much indeed. He
is no longer a doctor. He gave up practising when he left Sothernbay,
and accepted some sort of professorship—I forget what exactly; he
had to lecture on scientific subjects, that is about all I know. And he
writes a good deal too; he is becoming very well known—too well
known I fear; for I have heard it said that he is overworking himself
dreadfully. But he is not in England now; he is, I think, in India,
travelling with a party sent out by one of those learned societies, so I
fear my information isn’t of much use.”
“In India,” Cicely repeated sadly, but she said no more.
Soon after, Mr. Hayle left. But he went away, again promising to
call in a day or two.
“Has it tired you, mother?” said Cicely anxiously when he had
gone.
“No, dear. I like him; he is gentle and kind, and I like to see some
one who knew us at home,” she sighed a little. “At one time I shrank
from reviving any of the associations with our old life, but I am losing
that feeling.”
Cicely looked at her with wistful, anguished eyes. “Mother
dearest,” she said, “I don’t think I knew how much you loved
Greystone. Should I have kept it for you? Would you not have got ill
then? I might have hidden it all from you; Trevor would have been
glad to do so, and you need have never known there was any
trouble between us. Was I selfish and hasty—did I sacrifice you to
my pride? Have I done wrong, and is this my punishment?”
“My darling, no—a thousand times no,” replied her mother. “You
could not have married Trevor after—after his disappointing you so
terribly. You could not love where you did not respect. And you must
try to be more hopeful about me, dear, or I shall regret having told
you. I have felt better to-day than for some time past.”
Cicely tried to smile. “I think you do look a little better,” she said.
“Oh! I do so wish Mr. Guildford could have seen you.”
“He could not have done anything more. He said just the same as
Dr.——” replied Mrs. Methvyn. “My wishing to see him was more a
matter of feeling. He was so good to your father, I had learnt to trust
him; for your sake too, I wish he had been able to see me. He would
have understood it all so well.”
“Yes,” said Cicely. “I think he would.”
“He must be becoming quite a great man,” she went on after a
pause; “did you hear what Mr. Hayle said about him?”—“I dare say,”
she added to herself, “I dare say he has forgotten all about that fancy
of his. Men are better off than women; they can always bury trouble
in work.”
CHAPTER VII.
ALONE.
“Is Death as sad as Life?

Soon we shall know.
It does not seem to me
They find it so
Who die, and going from us
Smile as they go.”
Trefoil.
FOR some weeks Mrs. Methvyn seemed to gain ground, and

gradually, very gradually, Cicely’s fears abated. She began to think it
possible, or more than possible, that her mother had unconsciously
exaggerated her own danger; and that after all, many years of life
and, comparatively speaking, health, might be before her. And to this
opinion her mother’s greatly improved spirits seemed to lend a
strong colour of probability. The truth was, that Mrs. Methvyn felt
infinitely happier now that Cicely knew the worst; for it had been the
terror of breaking to her child these fresh tidings of impending woe,

PDF Relaxed Abduction Robust Information Interpretation For Industrial Applications 1St Edition Thomas Hubauer Auth Ebook Full Chapter

Uploaded by

Copyright:

Available Formats

You might also like

PDF Relaxed Abduction Robust Information Interpretation For Industrial Applications 1St Edition Thomas Hubauer Auth Ebook Full Chapter

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

PDF Relaxed Abduction Robust Information Interpretation For Industrial Applications 1St Edition Thomas Hubauer Auth Ebook Full Chapter

Uploaded by

Copyright:

Available Formats

Relaxed Abduction Robust Information

Interpretation for Industrial

Uncertainty Management for Robust Industrial Design in

Sensual Abduction Series Box Set Sensual Abduction 1 4

Computer Information Systems and Industrial Management:

Robust Control: Theory and Applications Kang-Zhi Liu

Industrial chemistry of oxides for emerging

Recombinant Microbes for Industrial and Agricultural

Industrial Enzyme Applications 1st Edition Andreas

Machine Learning Algorithms for Industrial Applications

Dissertation at the University of Lübeck, 2015

ISBN 978-3-658-14406-7 ISBN 978-3-658-14407-4 (eBook)

Library of Congress Control Number: 2016942010

Printed on acid-free paper

This Springer Vieweg imprint is published by Springer Nature

3.4 Extensions to the Basic Setting . . . . . . . . . . . . . . . . . 83

4 Case Studies and Evaluation 109

2.1 Pareto-optimality and Dominance . . . . . . . . . . . . . . . . 22

3.1 A conceptual model for diagnostics . . . . . . . . . . . . . . . 28

4.1 Relative runtime of RAPsolve-EL+ vs. RAPsolve-Generic . . 113

2.1 EL+ syntax and semantics . . . . . . . . . . . . . . . . . . . . 12

3.1 Relation between type of knowledge modelling, reasoning

© Springer Fachmedien Wiesbaden 2016

formal representation of a domain and an interpretation of information in

over imperfect domain formalisations that avoids the drawbacks outlined

1.2 Research Objectives

1.3 Scientiﬁc Contributions

industrial information interpretation. By formally relaxing the well-studied

1.4 Dissemination Activities

2.1 Ontologies and Description Logics

© Springer Fachmedien Wiesbaden 2016

interpretation function ·I that maps individuals to elements of the domain

Deﬁnition 2.1 (Logically independent axiom set)

Table 2.1: EL+ syntax and semantics

L we sometimes call ax an L-axiom. Analogously, a L-knowledge base is a

2.2 Abductive Reasoning

2.2.1 Standard Notions of Abduction

an inference scheme capable of providing possible explanations for some

• Knowledge-level abduction generalizes logic-based abduction with a

Deﬁnition 2.2 (Abduction problem)

6 Backward-chaining rule-based reasoning can obviously emulate this feature by breaking

Deﬁnition 2.3 (Preferential abduction problem)

2.2.2 Concept-based Notions of Abduction

Deﬁnition 2.4 (Subsumption-based abduction problem)

Lemma 2.1 (Axiom-based can simulate subsumption-based abd.)

Proof. Let H solve SAP, i. e. T |= Ci ∈H Ci  O. Then T ≡ T ∪

The notion of abduction introduced in Deﬁnition 2.2 is therefore non-

2.2.3 A Critique of Logic-Based Abduction

“It is not for what he would be to me now,

“Is Death as sad as Life?

FOR some weeks Mrs. Methvyn seemed to gain ground, and

You might also like

Proof. Let H solve SAP, i. e. T |= Ci ∈H Ci O. Then T ≡ T ∪