=Paper=
{{Paper
|id=Vol-2454/paper_37
|storemode=property
|title=Case-Based Retrieval and Adaptation of Regulatory Documents and their Context
|pdfUrl=https://ceur-ws.org/Vol-2454/paper_37.pdf
|volume=Vol-2454
|authors=Andreas Korger,Joachim Baumeister
|dblpUrl=https://dblp.org/rec/conf/lwa/KorgerB19
}}
==Case-Based Retrieval and Adaptation of Regulatory Documents and their Context==
https://ceur-ws.org/Vol-2454/paper_37.pdf
Case-Based Retrieval and Adaptation of
Regulatory Documents and their Context
Andreas Korger1 and Joachim Baumeister2,3
1
Angesagt GmbH, Dettelbachergasse 2, D-97070 Würzburg
2
denkbares GmbH, Friedrich-Bergius-Ring 15, D-97076 Würzburg
3
University of Würzburg, Am Hubland, D-97074 Würzburg
Abstract. Regulatory documents are required or provided by authori-
ties in many domains. They commonly point out relevant incidents for
specific scenarios. For those they have to present suitable preventive and
reactive measures. We introduce an approach to connect a case-based
description of the incidents structure with a case-based description of
the according context. This paper shows how to use case-based methods
to retrieve, adapt, and reuse incidents descriptions. Subsequently they
are used to generate new regulatory documents via case-based reasoning.
Case-based reasoning Experience Management Knowledge Management
SKOS Semantic Relatedness Natural Language Generation.
1 Introduction
A regulatory document describes incidents that are likely to happen in a cer-
tain situation. Preventive measures are elaborated to avoid the occurrence of
relevant incidents and adequate reactions are proposed. Further, harmful con-
sequences are to be avoided or mitigated. This underlying structure is repre-
sented in the documents structure. Popular examples of regulatory documents
are public events, for the handling of hazardous material or industrial workplace
safety. For a festival, a regulatory document would describe incidents such as
fire and relevant measures like the allocation of fire-extinguishers. The overall
goal of this work is to support domain experts in writing regulatory documents.
Fundamental considerations have been presented in preceding works [9, 10]. For
instance methods for documentary adaptation using a combination of ontolog-
ical document description and case-based reasoning. We extend ontologies to
represent these special parts of documents containing regulatory information.
Copyright c 2019 for this paper by its authors. Use permitted under Creative Com-
mons License Attribution 4.0 International (CC BY 4.0).
We use natural language processing to connect graph-based and textual knowl-
edge representation. Our goal is to retrieve and adapt passages of documents
depicting such regulatory knowledge for usage in another context.
For generating a new document, we need to answer three questions. Which
incidents are likely to happen? Which preventive and reactive measures are suit-
able for each incident under a certain context? How important is each measure?
The last question pays attention to the fact, that a limited budget and time
does not allow for the implementation of all preventive measures. This sums up
to consider a convenient context-based ranking for the incidents and measures.
The presented approach is a general framework and easily adaptable to domains
providing textual as well as ontological information for the context dependent
classification, prevention of, and reaction to incidents.
For reasons of simplification and consistency we give examples of the do-
main of public events. Our approach is driven by theoretical and case-based
considerations we describe in the first part of this work. Then we present the
experimental setup we used to install a case-study showing practical capabilities
of our approach. We finish with related work as well as with discussions and
future work.
2 Ontology Model for Incident Assessment
We make the assumption that there exists a corpus of regulatory documents
of a certain domain. The documents are sub-classified into passages, that are
connected with incidents or the according measures. Those passages of text are
called information units. The work of identifying these passages was done by do-
main experts. All information beyond the textual corpus of available documents
is coded into a knowledge base [3]. This knowledge base consists of entities (on-
tological concepts) as logical units and the relations between them. An entity
may be for instance an action, an agent, an event, or a resource. In a textual
context an entity is coded or described as one word (term) or more words up to
some sentences. Entities may be composed of sub parts in an arbitrary manner.
In the following, we introduce the basic concepts of our scenario.
Definition 1. Let KB = (E, R, D) be a knowledge base. Let E ⊂ KB be the set of
available entities (ontological concepts). Let I ⊂ E be the set of known incidents.
Let M ⊂ E be the set of known measures. Let R ⊆ E × E be the relations
between elements of E. Let D be the set of available documents D = {d1 , .., dj }.
Let U = {(u ∈ di )|di ∈ D} be the set of available information units contained in
D and T the set of terms used to textually build them.
It is very important for the assessment of safety measures to pay respect to
the context under that they are applied. For instance to supply rescue boats on
a festival F1 besides a river makes sense but for a festival F2 in the forest it
is totally senseless. The context are the factual parameters of the environment.
If the parameters change, the relevant incidents, the according measures and
the importances of both change. Respecting the documentary corpus D, the
context is for instance represented by certain parameters whose fulfillment is
mentioned in the content of each document or the parameters, all documents
have in common.
Definition 2. Let KB = (E, R, D) be the knowledge base. For an entity ei ∈ E
let Cei ⊆ E \ {ei } be the context of ei with Cei = {c1 , .., cj }.
For instance, for the previous entities F1 and F2 the context CF1 would be
near river and CF2 in the forest. We want to consider relations making enti-
ties a preventive or reactive measure to incidents. This means to focus on the
chronological order of the execution. In some domains measures are classified
into before, during and after an incident. We consider the relational classes dur-
ing and after as unified. A measure that is taken before an expected incident is
a preventive measure, a measure that is taken during or after an incident is a
reactive measure. A measure may be of preventive as well of reactive character.
Definition 3. Let RP M −C ⊆ M × I be a relation under a context C, indicating
which measures are taken in this context C before an incident, making them
preventive measures. Let RRM −C ⊆ I × M be the analogous relation, indicating
which measures are taken after the occurring of an incident making them reactive
measures.
Additionally we rely on an importance ranking of incidents and measures
under a given context. The importance is quantified by assigning a value between
1 for important and 0 for not important.
Definition 4. Let IM P (i, C) ∈ ]0, 1] be the importance of an element of I under
the context C. Let IM P (m, i, C) ∈ ]0, 1] be the importance of a measure m for
the incident i under the context C.
For a given context and relevant incident induced by the context the ac-
cording measures are ordered by importance and classified into preventive and
reactive measures. Altogether they build a kind of facilitated process snippet
we call PIRI (Preventive-Incident-Reactive-Interrelation). The presented model
simplifies the real world for facilitation of assessment. Typically there is a cascade
of measures that are executed in a specific order, e.g. in case of fire first evacuate
all people, then close the doors and windows. A PIRI-snippet is formally defined
as follows.
Definition 5. Let KB = (E, R, D) be the knowledge base. Let i ∈ I be an in-
cident and C ⊆ E a context. Then, we define a PIRI-snippet P IRI(i, C) =
{C, P G}, where PG = (N,E) is a directed graph. We call PG the PIRI graph
with nodes N = C ∩ M ∪ {i} all measures mentioned by C and the incident i
and edges E = {RP M −C ∪ RRM −C } ∩ {i} all edges containing the node i. The
graph is weighted with node-weights IMP(N,C) and edge weights IMP(E,i,C).
Figure 1 shows a PIRI-diagram for the incident I1 under the context C con-
sisting of several contextual elements (c1 , c2 , c3 , ..).
Incident
C1 P1 0,9 0,95
R1
Contextual
0,7
Elements
C2 Influence P2 0,85 I1
0,8
R2
0,7 0,75
C3 Preventive
Measures
P3 R3 Reactive
Measures
Fig. 1: PIRI-diagram under a context C = (C1 , ..., Cj ) showing the ranked pre-
ventive and reactive measures with the according importance weight.
2.1 Ontological Representation of the PIRI-Structure
Event base for Risk identify Incidents create Document
Classification Management Measures Structure
(ECLA) (PERM) (SECRI) (SECCO)
Fig. 2: Extension of the document creation workflow by the SECRI-ontology.
We extend a previously introduced ontology [9] by the definition of incidents,
measures and the PIRI-snippet. The existent ontology was used for the classifi-
cation of public events (OECLA ) and the structuring of the according regulatory
documents (OSECCO ) as depicted in Figure 2. For the ontological description of
incidents we now continued the elaboration of the SECRI ontology (OSECRI ).
The ontology describes the hierarchical context of incidents with a focus on pub-
lic events. We use the SKOS ontology [17] and the PROV ontology [11] as upper
ontologies. The SKOS ontology provides knowledge formalization and structur-
ing capability. The PROV ontology supports the representation of provenance
information to model the multi-agent-character of the scenario which is induced
by the involvement of several authors and addressees. For the ontological imple-
mentation of a PIRI-snippet we introduce the analogous classes and interweave
them with the documentary structure. An information unit is represented as a
secri:InformationUnit. This passage of text semantically targets a secri:Incident
or secri:Measure and is part of a document represented by secco:Document. In-
cidents and measures are subsuming classes as the top of a hierarchy. In the case
study we will see an example of this hierarchy in the domain of public events. A
graphical excerpt of the ontology can be seen in the Figure 3.
targets
partOf secri:InformationUnit
secco:Document
targets
secri:Incident targets
secri:Importance
targets targets
secri:Measure subClassOf
secri:PreventiveMeasure
subClassOf
secri:ReactiveMeasure
Fig. 3: Class representation of a PIRI-snippet.
2.2 Case-Based Representation of the PIRI-Structure
A convenient case-based representation for the so far described scenario inter-
nalizes the document description of incidents and measures. For each incident
mentioned by the regulatory document the preventive and reactive measures are
combined into a PIRI-snippet. We choose a structural case representation using
attributes and their values [5].
Definition 6. A case c1 = (d1 , l1 ) is defined by the incident and its context as
problem description d1 = {C1 , i1 } and its solution l1 = {(C1 ), (m ∈ RP M −C1 ∩
i1 ), (m ∈ RRM −C1 ∩ i1 )}, the combination of measures targeting the incident i1
under a context C1 , separated into preventive and reactive measures.
The problem descriptions and the solutions are conjunctions of elements of
the knowledge base. The context may be replaced for a unique identifier naming
the context without citing every component. For instance C1 = RegDocument1
then d1 = {RD1 , RainStorm} and l1 = {(RD1), (W eightT ents∧GetF orecast),
(CloseLiquidGas ∧ LockDoors ∧ GetRainCoat ∧ Evacuate)}.
Definition 7. The case base CB = {c1 , ..., cm } is the collection of all cases ci
extracted from available regulatory documents and constructed as described before
as PIRI-snippets. A query q to the case base is a conjunct subset of (negated)
measures and incidents.
For instance, the query q1 = CloseDoors ∧ ¬LockDoors ∧ Evacuation ∧
RainStorm retrieves all other PIRI-snippets containing an evacuation and a
closing and not locking of doors.
To retrieve cases, we search the case base for similar problem descriptions di
for the query q1 . To define a similarity function, we consider all preventive mea-
sures, reactive measures and the incident as individual sub-parts. Each of these
parts is then compared by a local similarity measure. With an aggregation func-
tion a global similarity measure is composed by weighting with the parameters
(ωP , ωI , ωR ) and summed up as follows:
SimPIRI (ck , cl ) = ωP SimP (Pk , Pl )+ωI SimI (ik , il )+ωR SimR (Rk , Rl ) /3 (1)
The incidents and measures are classified by a taxonomy that was derived from
the connected ontology, building the base for the similarity assessment and adap-
tation. The local similarities SimP , SimI , SimR are calculated via the taxonomic
order of its elements. The incidents I and the measures P,R are hierarchically
structured. Each element of the hierarchy is assigned with a likelihood symbol-
izing the similarity of its sub-elements. The similarity of the leaf-elements is set
to 1 and to 0 for the root element. The similarity increases with depth d of the
element according to for instance simd = 1−1/2d [1]. If we want to compare two
PIRI-snippets it is desirable to consider the context. For this reason we define
the following extended similarity measure under the context C:
SimContext (ck , cl ) = ω1 SimPIRI (ck , cl )+ω2 SimC (Contextk , Contextl ) /2 (2)
The context may for instance be the fulfillment of a classification hierarchy
describing the environmental parameters.
For instance, security measures under a context of high consumption of alco-
holic beverages are to be considered different as under a context of low consump-
tion of alcohol. So SimC is set to the similarity function used in that scenario
weighted by the weights ωi ∈ [0, 1] working as biases.
2.3 Constrained-based Extension
The importance ranking can also be used as an order of execution of measures.
The most important measures have to be taken first. But sometimes less impor-
tant measures have to be taken before other, more important measures. This
pays attention to the so called concatenation of circumstances. It is necessary
to introduce a (partial) order of measures additionally to the order induced by
preventive and reactive and the importance ranking.
Definition 8. For two measures m1 and m2 the constraint m1 ≺ m2 states that
m1 should be taken before m2 .
An obvious problem is as follows. To avoid theft or unauthorized access espe-
cially large buildings have to be locked after an evacuation. This can yield people
being locked inside the building. In reality it is often too complex or not possible
to describe for each incident an order of taking the measures. Additionally in a
multi-agent-scenario it is very difficult to execute instructions being too complex
or too numerous. We therefore take a simple strategy of providing only rules for
pairwise measures, as described before (Evacuate ≺ LockDoors).
3 Case Study
We exemplify the previous approach by a case study in the domain of public
events. We started with 15 regulatory documents in the domain of public events
that were annotated manually by three different domain experts. This corpus
is extended as a basis for the present evaluation. For the annotation process
we developed and evaluated several ontologies. These were used for the classifi-
cation of public events (OECLA ) and the structuring of the according security
documents (OSECCO ). The following Table 1 shows the number of ontological
concepts covered by each ontology.
Table 1. Number of ontological concepts for each ontology used.
OSECCO Structuring OECLA Classification OSECRI Incidents OSECRI Measures
278 136 115 72
3.1 Retrieval of Similar PIRI-Snippets
For the ontological description of security incidents we continue the elaboration
of the SECRI ontology (OSECRI ). The SECRI ontology describes the hierarchi-
cal context of security incidents for public events. An excerpt of the ontology can
be seen in Figure 4. In this work we extended the existing ontology by the ca-
pacity of modeling preventive and reactive measures for security incidents in the
domain of public events. All ontologies where implemented using the semantic
wiki KnowWE [4]. Amongst others we introduce the new classes secri:Measure
as well as secri:PreventiveMeasure and secri:ReactiveMeasure as subclasses of
secri:Measure.
secri:Disaggregation
broader
broader secri:ObeyAuthorities
broader
secri:Measure secri:CrowdControl broader
broader
broader secri:Reprimand
secri:InspectionMeasure secri:SiteDismissal
Fig. 4: Excerpt of the SECRI Ontology showing 7 of 72 measures.
For the case-based implementation we made use of myCBR [2]. The hierar-
chically structured incidents and measures represented in the SECRI ontology
were exported to a myCBR model. The case-based attributes were arranged into
taxonomies as local similarity measures. Those were aggregated into a global sim-
ilarity measure for the assessment of the according PIRI-snippets. A number of
relevant cases was extracted out of the corpus and installed in myCBR mak-
ing up the experimental case base. Table 2 shows the number of different cases
contained in the case base.
Table 2. Overview of different cases.
Different Contexts PIRI-Cases (Measures under Context)-Cases Information Units
15 300 1500 500
To evaluate the similarity assessment induced by the PIRI-strategy we con-
structed a post mortem analysis. This means to take every case of the case base
and use it as a query to the same case base. Our similarity measures are con-
structed symmetrically, consequently the query is commutative. As context we
use the event classification ontology Oecla . The context of each PIRI-snippet is
represented by the factual parameters classifying the event extracted out of the
according regulatory document. The pairwise similarities of the event classifica-
tion cases are already available due to a post mortem analysis done in previous
work [9]. Each document of the corpus mentions about 20 different incidents.
We now focus on the incident FireAndExplosion. For this incident we pairwise
calculate the similarity of the according PIRI-snippets. Afterwards we apply the
context and calculate the context dependent similarity. Figure 5 shows for each
pair of documents the similarity of the PIRI-snippet for the incident FireAnd-
Explosion as well as the PIRI-similarity combined with the context. This com-
parison makes clear, where the influence of the context changes the similarity
ranking of retrieved PIRI-cases.
3.2 Generation of Abstracted Information Units
In the following we present the strategy for the textual construction of PIRI-
snippets and their adaptation for reuse. Figure 6 shows the workflow of breaking
documents into reusable information units. Beginning with selected features it
shows how they can be put together to form a new document. It presents which
methods are used on each level for extraction, retrieval and adaptation. The
relevant textual elements were extracted from the corpus and transferred into the
ontological structures. The next step is to find the context dependent information
and replace it to make them reusable. We therefore searched for elements of the
domain vocabulary. Everything left we considered normal text or context related
information that can be abstracted. A strategy for abstraction is to replace words
by their class name. For instance a city name is replaced by location data or by
the part-of-speech class. The following exemplary text for the incident storm
shows, how an according passage of a security document would look in reality.
christm wine wine folk city carne folk music carne fair fair running camp arena campus
PIRI0 PIRI1 PIRI2 PIRI3 PIRI4 PIRI5 PIRI6 PIRI7 PIRI8 PIRI9 PIRI10 PIRI11 PIRI12 PIRI13 PIRI14
PIRI0 xx 0,5 0,6 0,6 0,6 0,5 0,5 0,6 0,6 0,5 0,5 0,3 0,5 0,3 0,5 0,4 0,5 0,3 0,5 0,3 0,5 0,3 0,3 0,3 0,3 0,4 0,5 0,0 0,2
PIRI1 0,5 0,7 xx 0,9 0,8 0,4 0,5 0,3 0,4 0,4 0,5 0,6 0,7 0,3 0,3 0,6 0,5 0,4 0,5 0,4 0,5 0,5 0,4 0,3 0,3 0,4 0,4 0,3 0,3
PIRI2 0,5 0,5 0,9 0,8 xx 0,5 0,5 0,4 0,4 0,5 0,4 0,6 0,6 0,3 0,3 0,5 0,5 0,4 0,6 0,4 0,5 0,4 0,4 0,3 0,4 0,4 0,4 0,2 0,3
PIRI3 0,5 0,5 0,4 0,5 0,5 0,5 xx 0,5 0,6 1,0 0,8 0,4 0,5 0,3 0,5 0,2 0,5 0,4 0,5 0,4 0,4 0,2 0,3 0,2 0,3 0,6 0,6 0,1 0,2
PIRI4 0,6 0,6 0,3 0,4 0,4 0,4 0,5 0,6 xx 0,5 0,6 0,3 0,4 0,2 0,5 0,1 0,4 0,1 0,3 0,1 0,2 0,2 0,3 0,1 0,2 0,7 0,6 0,0 0,1
PIRI5 0,5 0,5 0,4 0,5 0,5 0,4 1,0 0,8 0,5 0,6 xx 0,4 0,5 0,3 0,5 0,2 0,5 0,4 0,5 0,4 0,5 0,2 0,3 0,2 0,2 0,6 0,5 0,1 0,2
PIRI6 0,3 0,5 0,6 0,7 0,6 0,6 0,4 0,5 0,3 0,4 0,4 0,5 xx 0,3 0,4 0,3 0,5 0,5 0,5 0,3 0,4 0,3 0,3 0,3 0,3 0,4 0,5 0,1 0,3
PIRI7 0,3 0,5 0,3 0,3 0,3 0,3 0,3 0,5 0,2 0,5 0,3 0,5 0,3 0,4 xx 0,3 0,5 0,3 0,2 0,3 0,3 0,1 0,3 0,3 0,3 0,3 0,4 0,0 0,1
PIRI8 0,4 0,5 0,7 0,6 0,6 0,5 0,2 0,5 0,8 0,8 0,2 0,5 0,3 0,5 0,3 0,5 xx 0,5 0,5 0,5 0,5 0,6 0,5 0,4 0,4 0,2 0,3 0,1 0,1
PIRI9 0,3 0,5 0,4 0,5 0,4 0,6 0,4 0,5 0,1 0,3 0,4 0,5 0,5 0,6 0,3 0,2 0,5 0,5 xx 0,9 0,8 0,2 0,3 0,3 0,3 0,2 0,3 0,2 0,3
PIRI10 0,3 0,5 0,4 0,5 0,4 0,5 0,5 0,5 0,1 0,2 0,5 0,5 0,3 0,4 0,3 0,3 0,5 0,5 0,9 0,8 xx 0,2 0,3 0,3 0,3 0,2 0,3 0,2 0,3
PIRI11 0,3 0,3 0,5 0,4 0,5 0,4 0,2 0,3 0,2 0,3 0,2 0,3 0,3 0,3 0,1 0,3 0,6 0,5 0,2 0,3 0,2 0,3 xx 0,3 0,3 0,3 0,4 0,3 0,3
PIRI12 0,3 0,3 0,3 0,3 0,3 0,4 0,2 0,3 0,1 0,2 0,2 0,2 0,3 0,3 0,3 0,3 0,4 0,4 0,3 0,3 0,3 0,3 0,3 0,3 xx 0,1 0,2 0,0 0,3
PIRI13 0,4 0,5 0,4 0,4 0,4 0,4 0,6 0,6 0,7 0,6 0,6 0,5 0,4 0,5 0,3 0,4 0,2 0,3 0,2 0,3 0,2 0,3 0,3 0,4 0,1 0,2 xx 0,2 0,3
PIRI14 0,0 0,2 0,3 0,3 0,2 0,3 0,1 0,2 0,0 0,1 0,1 0,2 0,1 0,3 0,0 0,1 0,3 0,3 0,2 0,3 0,2 0,3 0,3 0,3 0,0 0,3 0,2 0,3 xx
Fig. 5: Post mortem analysis of the PIRI-snippets for the measure FireAndEx-
plosion without and with the context of the event classification. The values show
the similarities SimPIRI | SimContext . The value for SimContext was calculated out
of SimPIRI and SimECLA which were weighted with 0.5 each. A significant change
of the retrieval by the incorporated context is marked bold.
Corpus RD
Information Units
RD
RD
RD EXTRACT EXTRACT
RD1
RD
RD2 SELECT SELECT
ADAPT AND GENERATE ADAPT Ontological
Concepts
Document Similarity Assessment: Sentence Similarity Assessment: Concept Similarity Assessment:
ECLA, TF-IDF, SECRI PIRI, Sentence Embeddings Word embeddings, Ontologies,
Adaptation: User fills gaps Find and adapt similar information units Joint Embeddings
Fig. 6: Workflow of decomposing and recomposing security documents.
”Storm. Get weather information on a regularly basis from the munich weather
station. Weight all tents with heavy material or fix with ropes. In case of upcom-
ing storm, evacuate the event site using the franz josef avenue and call the fire
department.”
The PIRI-snippet with exemplary importance values for this would be:
Preventive(WeightTents(0.9),GetWeatherForecast(0.8))
Incident(Storm)
Reactive(CallFireDepartement(0.9),FullEvacuation(0.8)).
An abstracted information unit for the measure FullEvacuation would be:
”[FullEvacuation][StopWord][EventSite][Verb][StopWord][LocationData]”
This information unit can be adapted for instance to the measure Partial-
Evacuation. The ontological concept FullEvacuation is replaced by a retrieved
information unit for the new measure. The concept EventSite is for instance re-
placed by the more specific concept EventSiteComponent. This information can
be retrieved out of other cases because PartialEvacuation is commonly combined
with EventSiteComponent. The concept LocationData has to be replaced by the
contextual location information which is left to the user. The stop words are
inserted and corrected by a natural language generation tool or the user. The
generated textual passage before stop word correction and context correction
looks as follows:
”[Partial evacuation of the affected area][the]
[EventSiteComponent][using][the][LocationData]”
SELECT
NEW PROBLEM INITIAL FEATURES
1 3 4
Ontological 6
2 REUSE
Concepts QUERY ADAPTATION
OLD PROBLEM
RETRIEVAL
Multi Modal Knowledge Base
RD-NEW
CBR 5
RD-OLD Ontology Model
SD
SD
GENERATION Textual
SD
CBR
SD
SD
SD
SEMANTIC ANNOTATION
SD
Known
Corpus SOLUTION
Terms
7
RETAIN 8
Fig. 7: Case-based cycle of regulatory document assessment.
Figure 7 shows the user interaction and the case-based cycle of natural doc-
ument extraction and generation. In step (1) a new problem arises. That may
be for instance that a new regulatory document is required or an existing doc-
ument has to be improved as shown in step (2). All features are extracted out
of the problem description and the old document at step (3) and queried to the
knowledge base at step (4). The retrieved features, phrases and documents are
returned in step (5) and adapted in step (6) which requires user interaction. The
new regulatory document is used (7) and retained in the corpus enlarging the
case base (8).
3.3 Results and Discussion
The results of the case study for the retrieval of similar information units are
very promising even without user support. The incorporation of the contextual
paradigm significantly improved the simulation of the real world scenario. Re-
garding the generation of regulatory documents the results were quite good when
supported by the user. To answer the initial question, which incidents are likely
to happen, the context-based assessment can be used - similar context points to
similar incidents. Same holds for the measures that are suitable for an incident.
The importance of incidents and measures is made accessible by the percentage
of cases covering the incident or measure under a certain context.
4 Related Work
We started the research for related work to this paper with an overview of state
of the art publications in the domain of natural language generation presented
by Gatt and Krahmer [7]. Most of the presented work requires a large corpus for
the application of statistical methods. More suitable for our necessities seemed
grammar-based approaches. This lead us to the idea of abstracting text by giving
it a pseudo grammar structure.
There exists some work for the assessment of incidents in different domains.
A similar approach we want to mention was presented by Sizov et al. [14]. The
work focuses on the extraction and the (case-based) adaptation of explanations
contained in incident reports in the transportation domain. The work differs in
that way that we are aiming for a holistic document oriented and ontology-based
approach with user support for generation. A framework for the connection of
ontologies and constraints for the assessment of workflows was presented by
Nguyen et al. [12].
The structural integration of context into the case-based assessment was
covered by various authors. Different approaches for the incorporation of con-
text into a case-based decision were proposed by Pla et al. [13]. We adapted
the method of context stacking for this scenario. A conceptual revision of the
context-based reasoning paradigm was presented by Stensrud et al. [15]. For the
role of context in case-based reasoning a good work was published by Khan et
al. [8] as well as by Craw and Aamodt [6] for the use of similar case clusters
for representing context. The ontological side of context representation was for
instance presented in a thoughtful way by Strang et al. [16] and Xu et al. [18].
5 Conclusions
In this paper we presented a data structure called PIRI for the representation of
a regulatory document describing incidents and according measures. After for-
mally describing it, we transferred the structure into a case-based model. Using
this model an approach was shown for the adaptation of similarity measures to
different context. In a case study the approach was applied to a corpus of regu-
latory documents of the domain of public events. What we left for future work
are strategies for the identification of relevant attributes out of existing cases.
The application of attribute dependent weights would help to individually adjust
the influence of the context onto the case-based assessment. Additionally, in the
field of document generation the integration of grammar-based natural language
generation approaches seems to be promising. To adapt abstracted information
units to different contexts would help to reduce the needed user support.
References
1. Bach, K., Althoff, K.D.: Developing case-based reasoning applications using my-
CBR3. In: Agudo, B.D., Watson, I. (eds.) Case-Based Reasoning Research and
Development. pp. 17–31. Springer Berlin Heidelberg, Berlin, Heidelberg (2012)
2. Bach, K., Sauer, C., Althoff, K.D., Roth-Berghofer, T.: Knowledge modeling with
the open source tool myCBR. In: Proceedings of the 10th International Conference
on Knowledge Engineering and Software Engineering - Volume 1289. pp. 84–94.
KESE’14, CEUR-WS.org, Aachen, Germany (2014)
3. Baumeister, J., Reutelshoefer, J.: The connectivity of multi-modal knowledge
bases. CEUR Workshop Proceedings 1226, 287–298 (01 2014)
4. Baumeister, J., Reutelshoefer, J., Puppe, F.: KnowWE: A semantic wiki for knowl-
edge engineering. Applied Intelligence 35(3), 323–344 (2011)
5. Bergmann, R.: Experience Management. Springer, Berlin, Heidelberg (2002)
6. Craw, S., Aamodt, A.: Case-Based Reasoning as a Model for Cognitive Artificial
Intelligence: 26th International Conference, ICCBR 2018, Stockholm, Sweden, July
9-12, 2018, Proceedings, pp. 62–77 (07 2018)
7. Gatt, A., Krahmer, E.: Survey of the state of the art in natural language generation:
Core tasks, applications and evaluation. CoRR abs/1703.09902 (2017)
8. Khan, N., Alegre, U., Kramer, D., Augusto, J.C.: Is ‘context-aware reasoning =
case-based reasoning’ ? In: Brézillon, P., Turner, R., Penco, C. (eds.) Modeling and
Using Context. pp. 418–431. Springer International Publishing, Cham (2017)
9. Korger, A., Baumeister, J.: The SECCO ontology for the retrieval and generation
of security concepts. In: Cox, M.T., Funk, P., Begum, S. (eds.) ICCBR. Lecture
Notes in Computer Science, vol. 11156, pp. 186–201. Springer (2018)
10. Korger, A., Baumeister, J.: Textual case-based adaptation using semantic related-
ness - a case study in the domain of security documents. In: Wissensmanagement
Potsdam (2019)
11. Moreau, L., Groth, P.: Provenance: An Introduction to PROV. Synthesis Lectures
on the Semantic Web: Theory and Technology, Morgan and Claypool (2013)
12. Nguyen, T.H.H., Le-Thanh, N. (eds.): Ensuring the Semantic Correctness of Work-
flow Processes: An Ontological Approach. KESE 2014 Knowledge Engineering and
Software Engineering (2014)
13. Pla, A., Coll, J., Mordvaniuk, N., López, B.: Context-aware case-based reasoning.
In: Prasath, R., O’Reilly, P., Kathirvalavakumar, T. (eds.) Mining Intelligence and
Knowledge Exploration. pp. 229–238. Springer International Publishing, Cham
(2014)
14. Sizov, G., Ozturk, P., Marsi, E.: Let me explain: Adaptation of explanations ex-
tracted from incident reports. AI Communications 30, 1–14 (06 2017)
15. Stensrud, B.S., Barrett, G.C., Trinh, V.C., Gonzalez, A.J.: Context-based reason-
ing: A revised specification. In: FLAIRS Conference (2004)
16. Strang, T., Linnhoff-Popien, C., Frank, K.: Cool: A context ontology language to
enable contextual interoperability. vol. 2893, pp. 236–247 (01 2004)
17. W3C: SKOS Simple Knowledge Organization System Reference:
http://www.w3.org/TR/skos-reference (August 2009)
18. Xu, N., Zhang, W., Yang, H., Zhang, X., Xing, X.: Cacont: A ontology-based model
for context modeling and reasoning. Applied Mechanics and Materials 347-350 (03
2013)