Fuzzy Analogical Reasoning in Cognitive Cities

A Conceptual Framework for Urban Dialogue Systems

Stefan Markus Müller, Sara D’Onofrio and Edy Portmann

Human-IST Institute, University of Fribourg, Boulevard de Pérolles 90, Fribourg, Switzerland

Keywords: Cognitive City, Cognitive Computing, Computing with Words and Perceptions, Dialogue System, Fuzzy

Analogical Reasoning, Human Smart City, Soft Computing, Structure-Mapping Theory, Urban Governance.

Abstract: This article presents a conceptual framework for urban dialogue systems to let them emulate human analogical

reasoning by using cognitive computing and particularly soft computing. Since creating analogies is crucial

for humans to learn unknown concepts, this article proposes an approach of urban applications to human

cognition by introducing analogical reasoning as a sound component of their fuzzy reasoning process.

Pursuing an approach derived from (transdisciplinary) design science research, two experiments were

conducted to reinforce the theoretical foundation.

1 INTRODUCTION

Against the background of a strengthening

urbanization (e.g., dwindling urban living space) and

the increasing fuzziness in information (e.g., different

perceptions of a family-friendly neighborhood), cities

need to find new technological solution approaches to

manage plenty of data to counteract urban challenges,

such as natural resource use and human well-being.

Thereby, a promising approach is to enhance existing

urban systems with Web-based technologies (e.g.,

Web-of-things (D’Onofrio et al., 2018)) to sustain

urban governance and mainly increase efficiency and

sustainability, to finally establish smart cities

(D’Onofrio and Portmann, 2017).

Recent technological advances, such as sensor

technologies (Batty, 2013), have largely altered

characteristics of urban data (e.g., real-time instead of

past data). These advancements help to transform a

formerly sparse knowledge to a much more

sophisticated understanding of cities (Hurwitz et al.,

2015). Building upon incoming civic data (e.g.,

through citizens’ use of services), this understanding

is required to design and implement urban services

based on civic needs to shape human smart cities (i.e.,

becoming more human-oriented). By integrating

citizens as “drivers of change” into the development

of urban systems (e.g., civic tech movements), new

forms of participatory governance may arise (Oliveira

and Campolargo, 2015).

To foster information sharing in cities, existing

urban services can be provided with cognitive

capabilities, such as reasoning or learning abilities, to

mimic human intelligence (cf. humanistic computing

(Mann, 1998)). An approach, which might let urban

systems emulate human cognition, is the application

of cognitive computing. It enables supplementing

systems with cognitive processes, such as analogical

reasoning (Gentner, 1983), and, thus, accelerating

urban development to foster cognitive cities

(D’Onofrio and Portmann, 2017).

The authors present a conceptual framework for

urban dialogue systems to emulate human analogical

reasoning based on soft computing techniques (i.e., a

vital component of cognitive computing (D’Onofrio

and Portmann, 2017)), to suggest a new technological

solution approach (i.e., fuzzy analogical reasoning),

to improve existing urban systems with a focus on

socio-technical systems.

This article is an outline of a work-in-progress. To

this end, the authors use an approach derived from

design science research (Hevner and Chatterjee,

2010) that is advanced by transdisciplinary research

(Wickson et al., 2006) and follows the law of

parsimony (Laird, 1919). Section 2 presents the

theoretical background; the approach itself is outlined

in Section 3 and discussed in Section 4; Section 5

concludes this article.

Müller, S., D’Onofrio, S. and Portmann, E.

Fuzzy Analogical Reasoning in Cognitive Cities.

DOI: 10.5220/0006816103530360

In Proceedings of the 20th International Conference on Enterprise Information Systems (ICEIS 2018), pages 353-360

ISBN: 978-989-758-298-1

353

2 BACKGROUND

By outlining how human smart cities may advance to

cognitive cities, describing soft computing as a vital

component of cognitive computing, and introducing

structure-mapping theory, this section presents the

theoretical foundation of the framework.

2.1 The Role of Technology and

Humans in Urban Development

Building upon advanced Web-based technologies

(e.g., Web-of-things (D’Onofrio et al., 2018)), smart

urban systems, such as dialogue systems, represent

potential starting points for enriching civic

interaction. They enable the exchange of citizens’

perceptions and knowledge among them and foster

urban development, particularly regarding efficiency

and sustainability. Moreover, they allow to build a

collective knowledge base, supporting urban

decisions on a data-driven basis (Finger and

Portmann, 2016; Malone and Bernstein, 2015).

Based on aggregated data sets obtained through

citizens’ use of urban services (e.g., everyday

questions through civic interaction), cities obtain an

integrated view on issues (e.g., urban living space)

and can involve affected stakeholders (e.g., citizens)

specifically (Hurwitz et al., 2015). Hence, cities can

take broad-based decisions to improve equity and

sustainability of urban life. Thus, participative

models of urban governance are established, allowing

the development of human smart cities by putting

citizens in foreground and giving them the possibility

to shape their living environment through the

expression of their needs or ideas for improvements

(Oliveira and Campolargo, 2015).

Human smart cities can be further reinforced by

being supplemented with cognitive computing.

Cognitive cities build upon cognitive systems and

processes (e.g., reasoning or learning processes) and

are increasingly capable of dealing with a human

living environment that is constantly changing and

getting more complex (Mostashari et al., 2011). Due

to natural language, which is seen as the main

communication medium in cities, urban environment

is also getting fuzzier. Therefore, cognitive cities

enable developing collective and humanistic

intelligence (i.e., citizens and urban systems are

working together using natural language) (Malone

and Bernstein, 2015; Mann, 1998), which

significantly addresses urban resilience by helping to

encounter challenges, such as urbanization and

digitalization (D’Onofrio and Portmann, 2017).

2.2 Soft Computing

Fuzzy logic represents an extension of traditional

logic where p can be true or false or have an

intermediate truth value (Zadeh, 1988). This allows a

generalization of conventional set theory, namely

fuzzy set theory, where an element x of a finite or

infinite set X is no longer either contained or not in a

crisp subset A in X. Instead, it is possible to define a

membership function µ

(x), which shows to what

degree an element x is contained in a fuzzy subset A

in X (Zadeh, 1965):

(x) → [0, 1]

(1)

To build fuzzy sets, it is required that humans can

break down information into various levels of

abstraction by reflecting hierarchical structures of

their environment upon the hierarchical organization

of their knowledge (Yao, 2006). Thereby, the ability

to decompose the complex and uncertain

environment into simpler and tractable clumps (i.e.,

fuzzy information granules (Zadeh, 1997)) is crucial

for human information processing (Hobbs, 1985). For

instance, general words represent high levels of

granularity, while specific words express high levels

of detail, but low levels of granularity (Yao, 2006).

Hence, fuzzy information granulation helps citizens

to wrap up data and reduce information overload.

By using fuzzy information granules, computing

with words (CWW) allows to describe human-like

reasoning based on fuzzy logic (D’Onofrio and

Portmann, 2015). Since most real-world constraints

have a tolerance for fuzziness, the concept of a

generalized constraint (GC) strives to define degrees

of fuzziness based on CWW (Zadeh, 2005):

GC: X isr R

(2)

Thereby, X is a constrained variable and R a fuzzy

constraining relation. The modularity (i.e., semantics)

of R is identified by r, an indexing variable that is

adjustable (e.g., equal, possibilistic, probalistic)

(Zadeh, 2005). Assuming possibilistic semantics of

R, r is abbreviated to a blank space and a GC is

adjusted as follows (Zadeh, 1996):

GC: X is R

(3)

With possibilistic semantics, R constrains a

variable X by playing the role of the possibility

distribution of X. If u is a generic value of X, and µ

is a membership function in R, the semantics of R can

be defined as follows (Zadeh, 1996):

Poss{X = u} = µ

(u)

(4)

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

354

Hence, methods based on fuzzy logic foster

human-like reasoning and facilitate, among others,

analogical reasoning.

2.3 Structure-Mapping Theory

Analogy represents abstractions of higher-order

human cognition (i.e., symbolic information

processing), which involves humans mapping one

situation onto another to process information, learn

and acquire new knowledge. In the case of analogy,

humans focus on relational similarity (i.e., structure-

based similarity) between two situations, which is

based on labelled relations (i.e., symbols) (Boteanu

and Chernova, 2015). Hence, an analogy is apparent

if a relation between a pair of data elements d and e is

structurally similar to a relation between another pair

of data elements g and h (e.g., district is to city as

chapter is to book) (Barr et al., 2015).

Structure-mapping theory (SMT) explains not

only analogical but also similarity reasoning and,

thus, how humans map data elements of a familiar

situation (i.e., base) onto data elements of an

unfamiliar situation (i.e., target) to understand and

draw new inferences about the latter. Each mapping

comprises a set of correspondences either between

attributes (i.e., features) of data elements or relations

(i.e., structures) among one another. Depending on

whether correspondences between data elements

emphasize their attributes (i.e., objective similarity)

or their relations among one another (i.e., relational

similarity), different types of similarity (e.g., surface

similarity, analogy) present the outcome of the

mapping of two situations (Gentner, 1983).

Based on common patterns of entities (i.e.,

attributes, relations), which have emerged from the

alignment of two situations, humans make new

presumptions (i.e., candidate inferences) about the

target. Pursuing structural consistency and

systematicity, humans complete these patterns and

familiarize with unknown data elements. An analogy

occurs if and only if humans familiarize with a target

by drawing inferences based on relational pattern

completion (Gentner and Markman, 1997).

3 CONCEPTUAL FRAMEWORK

This section presents related work, the concept and

the evaluation of the proposed framework for urban

dialogue systems to emulate human analogical

reasoning drawing on soft computing techniques.

3.1 Related Work

Dialogue systems aim to automatically deliver

relevant and concise answers to humans’ questions

often posed in natural language (Ojokoh and

Ayokunle, 2013). Such systems are of importance as

they can help to sustain urban governance by

enhancing the information exchange between cities

and citizens. Since natural language consists of

linguistic variables (e.g., words), the inclusion of soft

computing techniques may improve dialogue systems

(Zadeh, 2006). Although there have been

developments of such reasoning methodologies (e.g.,

CWW-based system (Khorasani et al., 2009),

perception-based system (Ahmad and Rahimi,

2006)), these applications, such as expert systems,

typically do not foster reasoning and dialogue in such

a way that humans process information by using, for

instance, analogical reasoning (Zadeh, 2006).

Moreover, there have been aspirations to facilitate

natural language processing (NLP) by means of

analogical reasoning (e.g., denominal verb

interpretation (McFate and Forbus, 2016), word sense

disambiguation (Barbella and Forbus, 2013)).

However, none of these systems constitute an

approach to human cognition because attempts, such

as NLP, do not account for fuzziness in the

characterization of biological systems (Seising and

Sanz, 2012). Since it is proposed that cognitive

systems use soft computing techniques to become

capable of understanding and extracting

heterogeneously structured information from natural

language (Zadeh, 2005), their ability to create

analogies should also consider dealing with fuzzy

information.

3.2 Concept

Based on previous work of Bouchon-Meunier and

colleagues (Bouchon-Meunier et al., 2003; Bouchon-

Meunier and Valverde, 1999), the proposed

framework builds upon an analogical scheme for

approximate reasoning to allow urban systems to

interact with citizens in natural language. Starting

with a citizen’s question, for example “How can I get

a seasonal-work approval?”, the system decomposes

it into data elements consisting of words or a

sequence of words: how, get, seasonal-work and

approval. By completing the granulation process (cf.

Zadeh, 1997), the system needs to clarify through

alignment with existing stored knowledge whether it

understands single data elements. In this example, the

following classification is possible: how belongs to

certain degrees to the fuzzy set factoid question and

Fuzzy Analogical Reasoning in Cognitive Cities

355

that of procedure, get to the fuzzy set infinitive, and

seasonal-work and approval remain unknown.

Resulting from the system’s granulation process,

two linguistic variables X and Y are assumed for the

unknown information: X represents a domain with

possible values p (e.g., temporary-work) and q (e.g.,

seasonal work), and Y another domain that can take

values r (e.g., permit) and s (e.g., approval).

Assuming furthermore possibilistic semantics of a

fuzzy constraining relation R

, the system can

formulate a GC for Y (Zadeh, 1996):

GC: Y is R

where Poss{Y = r} = µ

(r)

and Poss{Y = s} = µ

(s)

(5)

Next, the system may use resemblance relations

(cf. Bouchon-Meunier and Valverde, 1999) to gain a

somewhat known value for q based on a known

relation R

between the unknown data element q and

a known data element p. Having retrieved a relation

between linguistic values p and q, the system

becomes capable of drawing the analogical scheme to

gain a known value for an unknown data element s

(Bouchon-Meunier et al., 2003):

Figure 1: Analogical scheme.

Therefore, the system can use a fuzzy-based

application of the compositional rule of inference

(CRI) (i.e., approximate reasoning). The CRI is an

extension of the familiar rule of inference (i.e.,

generalized modus ponens), which states if X is true

and implies Y, then Y is true, and is applied as follows:

(r) = max

(µ

(p) ∧ µ

βRXRY

(p, r))

(6)

Based on resemblance relations (i.e., p and q are

apparently known to be related by R

), temporary-

work and seasonal-work are approximately equal and,

thus, seasonal-work is more or less permit.

Since the linkage β between p and r is also known

(i.e., temporary-work belongs to the fuzzy set permit

to a not negligible degree), it is possible to gain a

known value for s, which is to q as r is to p, drawing

an inference R

between data elements r and s

(Bouchon-Meunier and Valverde, 1999). In terms of

SMT, a structure-based correspondence is projected

based on β (Gentner, 1983). Therefore, the system can

project a mapping R

βRXRY

, which provides it with a

linguistic value for s based on relations β, R

, R

expressed by their membership grades in respective

sets of fuzzy (sub-)sets of X and Y (i.e., μ

, μ

)

(Bouchon-Meunier et al., 2003):

βRXRY

: μ

(p) × μ

(r) × μ

(q) → μ

(s),

where β ⊂ μ

(p) × μ

(r)

and R

⊂ μ

(p) × μ

(q)

and R

⊂ μ

(r) × μ

(s)

(7)

Assuming that approval is to seasonal-work as

permit is to temporary-work, the system creates a

mapping to understand approval in relation to

seasonal-work. The system ends the analogical

reasoning process by classifying all data elements as

known and may continue the dialogue with the

citizen.

3.3 Evaluation

This subsection briefly outlines how the conceptual

framework was evaluated through two experiments,

both carried out by the authors, pursuing a

methodological approach oriented towards design

science research in information systems and

transdisciplinary research to include citizens into the

development process (Hevner and Chatterjee, 2010).

3.3.1 Workshop-based Experiment

In May 2017, the authors conducted a workshop to

become more familiar with citizens’ requirements for

new (smart) urban systems and, simultaneously, get

insights about the human reasoning process. Nine

males and two females, all between 20 and 50 years

old and with different professional backgrounds (e.g.,

computer science, geography), participated in the

workshop. It lasted two hours in total, whereby the

experiment took half an hour.

First, an introduction about cognitive computing

and soft computing was given to build a theoretical

foundation for the experiment. Afterwards, a

discussion about the term human-machine interaction

began, exchanging opportunities and threats of

computer systems that are able to compute natural

language (e.g., Alexa, Siri). Most participants were

sceptical about systems that are taking over control or

being able to autonomously make decisions.

However, some participants stated that they are

curious to test such “intelligent” systems (e.g., self-

driving cars). They even considered using them in the

future if such services would turn out to be

advantageous for them.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

356

After having discussed smart systems and, more

concretely, their acceptance and usefulness for cities,

the main experiment was conducted with the aim to

reinforce the theories about the relation between

questions and answers as well as to get constructive

inputs regarding the human reasoning process. The

experiment consisted of two parts.

The first part was about answering five “W”-

questions (i.e., who, where, when, how long, how)

stated in German. The results showed that almost

every participant answered in the same way, even if

the answers were not identically (e.g., whole name vs.

last name). To grasp the essence of the results, some

examples are presented: To the question “Who leads

the experiment?”, everyone answered with a name,

obviously with the one of the moderating author.

Asking “Where does the experiment take place”, all

participants stated a location. Their responses were

only differing in their granularity (e.g., Bern vs.

Impact Hub Bern). To the question “How long does

the workshop take?”, everyone answered with a time

specification, however, also containing granular

variances, such as mentioning hours or minutes.

Considering the semantics of all answers, this

experiment showed that “W”-questions influence

humans in the way of how they give responses.

The second part consisted of responding to three

questions that were impossible to answer because of

their semantics (e.g., “Where do lucky devils grow?”).

Even if the “W”-word provided hints on how to

answer the question, no meaningful linking with

existing knowledge was possible. Therefore, the

participants were not able to give a reasonable

answer. They tried to create analogies using familiar

situations and, thus, responded with information that

most likely matched them individually. Although the

questions were obviously not meant serious, no one

considered the possibility of stating that there is no

answer (even though that would have been the correct

answer) and instead gave unserious answers (e.g.,

haunted forest).

3.3.2 Laboratory Experiment

In May 2017, a laboratory experiment was conducted

to evaluate the stableness of the conceptual

framework’s theoretical foundation. Its purpose was

to document how far subjects would follow

theoretical predictions of SMT if they created

analogies to find relations between their existing

knowledge and unknown concepts.

Seven males and three females, all between 25

and 30 years old, participated in the experiment,

where everybody was of Swiss nationality and had an

academic background. The paper-pencil experiment

lasted 20 minutes in total and pursued double-blind

anonymity as well as a 1x2 between-subject design.

The two treatments split into an experimental

treatment and a control treatment, completed by five

subjects each.

SMT predicts that objective similarity is typically

more likely to be retrieved by humans than relational

similarity because it is represented by superficialities,

which are easier to recollect (Gentner and Forbus,

2011). Hence, the tested hypotheses postulated this

fundamental assumption and are outlined as follows:

H1: If humans need to retrieve a familiar situation

(i.e., base) by their own memory to understand and

draw new inferences about an unfamiliar situation

(i.e., target), they tend to encode objective similarity

(i.e., surface similarity) rather than relational

similarity (i.e., analogy).

H2: If an unfamiliar situation (i.e., target)

immediately comes with a familiar situation (i.e.,

base), which can be used to understand and draw new

inferences about the former (e.g., analogical

argumentation in a discussion), humans tend to

encode relational similarity (i.e., analogy) rather

than objective similarity (i.e., surface similarity).

The questionnaire used was in German, consisted

of five tasks in two variations: One that was

completed by the experimental treatment and another

one by the control treatment. Thereby, the treatment

variable described whether subjects needed to

retrieve a familiar situation by their own memory to

map any form of similarity onto an unfamiliar

situation. Hence, the treatment variable was

nominally scaled and took the values 0 (i.e., subject

was in the control treatment) or 1 (i.e., subject was in

the experimental treatment). Two graphical

representations were used to illustrate such familiar

or unfamiliar situations.

For both treatments, it was measured afterwards

which form of similarity subjects had tended to

encode either through their choices (i.e., control

treatment) or their own drawings (i.e., experimental

treatment). The dependent variable here was

nominally scaled as well and took the values 0 (i.e.,

relational similarity) or 1 (i.e., objective similarity).

These measurements served to investigate the

statistical correlations and conditional probability

distributions between the two elicited variables.

In H1, the descriptive univariate analysis

indicated that subjects in the experimental treatment

encoded relational and objective similarity equally.

Although they needed to retrieve a base by their own

memory, subjects did not tend to draw graphical

representations that primarily shared an objective

Fuzzy Analogical Reasoning in Cognitive Cities

357

similarity with the target. In H2, conditional

probability distributions of the control treatment

suggested that subjects clearly tended to encode

relational similarity rather than objective similarity,

as they were given both target and base

simultaneously. Finally, in H1 and H2, the descriptive

bivariate data analysis indicated that there was a

moderate correlation between the form of similarity,

which subjects tended to encode, and whether they

needed to retrieve a familiar situation by themselves

to map this similarity onto an unfamiliar situation.

4 DISCUSSION

The conceptual framework builds upon an analogical

scheme for approximate reasoning, which denotes a

type of reasoning that is neither quite precise nor quite

imprecise and expresses humans’ ability to take

rational decisions in complex and uncertain

environments. By applying the analogical scheme and

linking analogical concepts to soft computing, (Web-

based) urban dialogue systems might understand

previously unknown data elements. Hence, if they do

not understand one or several data elements during an

interaction with a citizen, they may create an analogy

to find a relation with existing stored knowledge

(Bouchon-Meunier and Valverde, 1999). Since

creating analogies is crucial for humans to learn

unknown concepts and soft computing allows to

understand and extract information from natural

language, urban dialogue systems would become

more oriented towards humans and perform and learn

better (D’Onofrio and Portmann, 2017; Gentner,

2010).

To reinforce the theories, two experiments were

conducted (independently from each other). In the

workshop-based experiment, the focus was put on

question-answering processes to get constructive

inputs regarding the human reasoning process.

Thereby, two valuable findings were gained: First,

“W”-questions influenced the way of how to answer

(e.g., who = person, where = location), and second, if

the semantics of a question made no or little sense,

participants tended to create analogies using familiar

situations. As an illustration, the question in German

“Wo wachsen Glückspilze?” (engl. “Where do lucky

devils grow?”) is presented here: Some participants

mentioned to have tried to create analogies using Pilz

(engl. mushroom). Being influenced by the question

word wo (engl. where), participants searched for a

possible place (e.g., haunted forest) as an answer,

associating Glück (engl. luck) with a magical element.

They largely justified their answers with an

association to their childhood, in which they got to

know fairy tales. Thus, they connected the unknown

word (i.e., the growth place of lucky devils) with an

element known from their knowledge base. Hence,

creating analogies is crucial for human information

processing and, therefore, for future urban dialogue

systems, too.

In the laboratory experiment, the results mostly

indicated a stable theoretical foundation of the

developing reasoning process. However, not all tested

hypotheses were provided with evidence in favour of

SMT. The most interesting finding was that relational

similarity (i.e., analogy) was equally encoded by

subjects, even if they had completed a retrieval by

themselves. Hence, support for the hypothesis behind

that finding (i.e., H1) could not be drawn, which

counters a theoretical prediction of SMT. Regarding

dialogue systems emulating human analogical

reasoning, this last finding contains a promising

implication: Since systems at some point need to

retrieve a relation based on the drawn analogical

scheme, it would further improve urban services and

be in favour of citizens, if systems were able to

encode powerful relational similarity by default and

did not provide citizens with answers based on

objective similarity (Gentner et al., 1993).

By further emulating human information

processing through analogical reasoning, cognitive

systems might perform better in interaction with

citizens, make human-computer interaction (HCI)

even more human-centered and facilitate the urban

learning process additionally (Gentner, 2010). This

provides the foundation for a resilient urban network

of knowledge that is driven by a constantly learning

collective intelligence (Malone and Bernstein, 2015).

Thereby, intelligence amplification denotes a

visionary concept that outlines how continuous and

complementary HCI may shape and augment urban

intelligence and, thus, sustain society (D’Onofrio and

Portmann, 2017). Shaping and increasing urban

intelligence are not simple endeavors. This is because

amounts of data storage, communication capacity

and, hence, potential knowledge for humans are

exponentially growing, human information

processing however remains unchanged (Batty,

2013). Therefore, human-centered, mutual and

constant HCI is increasingly necessary to enhance

human reasoning and learning.

If cities integrate cognitive computing into urban

applications that may also emulate human analogical

reasoning, it is crucial that they pursue a business and

an economic plan that ensure appropriate planning,

procurement and delivery of corresponding

technologies and infrastructures. Thus, an appropriate

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

358

environment for a human-computer symbiosis can be

established and collective and humanistic intelligence

sustainably be created (Malone and Bernstein, 2015;

Mann, 1998). This intelligence is needed eventually

to strengthen urban resilience and sustain urban

governance to tackle challenges of urbanization and

digitalization (Finger and Portmann, 2016).

5 LIMITATIONS AND OUTLOOK

The presented framework for urban dialogue systems,

which is conceptually designed in this article,

represents an extract of a developing idea of a global

reasoning process for urban systems and is part of a

current research project. Therefore, most limitations

and corresponding suggestions for future research

focus on this research project.

More research needs to be done relating to how a

system decomposes a citizen question and whether

this is expedient at all. So far, data elements have

received a meaning only after they were decomposed

by granular computing into words or a sequence of

words. However, data elements might have a different

meaning if they are processed as part of an entire

question (Chowdhury, 2003). A conductive thought

can be that the system immediately tries to align the

citizen question with existing stored knowledge,

looking for similar questions that have been asked in

the past and might help to classify the incoming one.

Furthermore, investigation of modern information

retrieval (IR) techniques is needed. This is because

techniques based on conventional models are not in

favour of the IR process. Relating to a fuzzy IR

process, there are promising approaches that might be

specified for the proposed fuzzy reasoning process

(Baeza-Yates and Ribeiro-Neto, 2011).

More specifications are necessary regarding the

nature of a system’s knowledge base. Known fuzzy

sets first need to be collected and stored such that they

can be associated with processed data elements

afterwards. This brings up the question about an

appropriate computer memory. The authors intend to

use fuzzy cognitive maps (FCMs) as a memory basis

stored in graph databases (cf. D’Onofrio et al., 2017).

Since it would define relevance gradually, the system

would answer questions effectively even if they were

formulated imprecisely. Therefore, an alignment of

FCMs with SMT might be an expedient next step.

Finally, it needs to be noted that this introduction

of analogical reasoning as a sound component of the

fuzzy reasoning process focuses on SMT because it is

displayed in the analogical scheme, which links

analogical reasoning to soft computing. However,

there are several analogical concepts (e.g., metaphor,

schema, transfer) that might also provide a basis for

further development of the fuzzy reasoning process.

One last limitation relates to both experiments

whose findings do not raise a claim to represent high-

level scientific contributions, particularly not in

methodological terms. The experiments served much

more as an evaluation of the developing fuzzy

reasoning process as a designed artefact, which is

oriented towards the process of (transdisciplinary)

design science research. The authors encourage

further urban researchers to conduct experiments with

citizens to grasp actual existing needs and, therefore,

to develop meaningful urban systems.

ACKNOWLEDGEMENTS

The authors would like to thank the participants of the

workshop as well as the participants of the laboratory

experiment for the valuable insights.

REFERENCES

Ahmad, R., Rahimi, S., 2006. A Perception Based, Domain

Specific Expert System for Question-Answering

Support. In IEEE/WIC/ACM International Conference

on Web Intelligence. Hong Kong: IEEE/WIC/ACM,

893-896.

Baeza-Yates, R., Ribeiro-Neto, B., 2011. Modern

Information Retrieval. New York: ACM Press.

Barbella, D., Forbus, K.D., 2013. Analogical Word Sense

Disambiguation. In Advances in Cognitive Systems,

2(1): 297-315.

Barr, N., Pennycook, G., Stolz, J.A., Fugelsang, J.A., 2015.

Reasoned connections: A dual-process perspective on

creative thought. In Thinking and Reasoning, 21(1): 61-

75.

Batty, M., 2013. Big data, smart cities and city planning. In

Dialogues in Human Geography, 3(3): 274-279.

Boteanu, A., Chernova, S., 2015. Solving and Explaining

Analogy Questions Using Semantic Networks. In

Proceedings of the Twenty-Ninth AAAI Conference on

Artificial Intelligence. Austin: AAAI, 1460-1466.

Bouchon-Meunier, B., Mesiar, R., Rifqi, M., 2003.

Compositional rule of inference as an analogical

scheme. In Fuzzy Sets and Systems, 138(1): 53-65.

Bouchon-Meunier, B., Valverde, L., 1999. A fuzzy

approach to analogical reasoning. In Soft Computing,

3(1): 141-147.

Chowdhury, G.G., 2003. Natural Language Processing. In

Annual Review of Information Science and Technology,

37(1): 51-89.

D’Onofrio, S., Franzelli, S., Portmann, E., 2018. Advancing

Cognitive Cities with the Web of Things. In Yager,

R.R., Espada, J.P., eds., New Advances in the Internet

Fuzzy Analogical Reasoning in Cognitive Cities

359

of Things, 1

ed. New York: Springer Science and

Business Media, 75-91.

D’Onofrio, S., Portmann, E., 2015. Von Fuzzy-Sets zu

Computing-with-Words. In Informatik Spektrum,

38(6): 543-549.

D’Onofrio, S., Portmann, E., 2017. Cognitive Computing in

Smart Cities. In Informatik Spektrum, 40(1): 46-57.

D’Onofrio, S., Wehrle, M., Portmann, E., Myrach, T., 2017.

Striving for Semantic Convergence with Fuzzy

Cognitive Maps and Graph Databases. In International

Conference on Fuzzy Systems. Naples: FUZZ-IEEE, 1-

Finger, M., Portmann, E., 2016. What Are Cognitive

Cities? In Portmann E., Finger, M., eds., Towards

Cognitive Cities: Advances in Cognitive Computing

and its Applications to the Governance of Large Urban

Systems, 1

ed. Cham: Springer International

Publishing, 1-11.

Gentner, D., 1983. Structure-Mapping: A Theoretical

Framework for Analogy. In Cognitive Science, 7(1):

155-170.

Gentner, D., 2010. Bootstrapping the Mind: Analogical

Processes and Symbol Systems. In Cognitive Science,

34(5): 552-775.

Gentner, D., Forbus, K.D., 2011. Computational models of

analogy. In Cognitive Science, 2(3): 266-276.

Gentner, D., Markman, A.B., 1997. Structure Mapping in

Analogy and Similarity. In American Psychologist,

52(1): 45-56.

Gentner, D., Rattermann, M.J., Forbus, K.D., 1993. The

Roles of Similarity in Transfer: Separating

Retrievability from Inferential Soundness. In Cognitive

Psychology, 25(4): 524-575.

Hevner, A., Chatterjee, S., 2010. Design Science Research

in Information Systems. In Hevner A., Chatterjee, S.,

eds., Design Research in Information Systems, 1

ed.

Boston: Springer, 9-22.

Hobbs, J.R., 1985. Granularity. In International Joint

Conference on Artificial Intelligence. Los Angeles:

IJCAI, 432-435.

Hurwitz, J., Kaufman, M., Bowles, A., 2015. In Cognitive

Computing and Big Data Analytics, 1

ed. Indianapolis:

John Wiley and Sons.

Khorasani, E.S., Rahimi, S., Gupta, B., 2009. A Reasoning

Methodology for CW-Based Question Answering

Systems. In Di Gesù V., Pal S.K., Petrosino, A., eds.,

Fuzzy Logic and Applications, 1

ed. Berlin,

Heidelberg: Springer, 328-335.

Laird, J., 1919. The Law of Parsimony. In The Monist,

29(3): 321-344.

Malone, T.W., Bernstein, M.S., 2015. Handbook of

Collective Intelligence, 1

ed. Cambridge: MIT Press.

Mann, S., 1998. Humanistic Computing: “WearComp” as a

New Framework and Application for Intelligent Signal

Processing. Proceedings of the IEEE, 86(11): 2123-

2151.

McFate, C., Forbus, K.D., 2016. Analogical Generalization

and Retrieval for Denominal Verb Interpretation. In

Conference of the Cognitive Science Society. Austin:

Cognitive Science Society, 1277-1282.

Mostashari, A., Arnold, F., Mansouri, M., Finger, M., 2011.

Cognitive cities and intelligent urban governance. In

Network Industries Quarterly, 13(3): 4-7.

Ojokoh, B., Ayokunle, P., 2013. Online Question

Answering System. In International Journal of

Computer Science Research and Application, 3(3): 2-9.

Oliveira, A., Campolargo, M., 2015. From Smart Cities to

Human Smart Cities. In 48th Hawaii International

Conference on System Sciences. Kauai: HICSS, 2336-

2344.

Seising, R., Sanz, V., 2012. From Hard Science and

Computing to Soft Science and Computing – An

Introductory Survey. In Seising R., Sanz, V., eds., Soft

Computing in Humanities and Social Sciences, 1

ed.

Berlin, Heidelberg: Springer, 3-36.

Wickson, F., Carew, A.L., Russell, A.W., 2006.

Transdisciplinary research: characteristics, quandaries

and quality. In Futures, 38(9): 1046-1059.

Yao, Y.Y., 2006. Three Perspectives of Granular

Computing. In Journal of Nanchang Institute of

Technology, 25(2): 16-21.

Zadeh, L.A., 1965. Fuzzy Sets. In Information and Control,

8(1): 338-353.

Zadeh, L.A., 1988. Fuzzy Logic. In IEEE Computer, 21(4):

83-93.

Zadeh, L.A., 1996. Fuzzy Logic = Computing with Words.

In IEEE Transactions on Fuzzy Systems, 4(2): 103-111.

Zadeh, L.A., 1997. Toward a theory of fuzzy information

granulation and its centrality in human reasoning and

fuzzy logic. In Fuzzy Sets and Systems, 90(2): 111-127.

Zadeh, L.A., 2005. Toward a generalized theory of

uncertainty (GTU) – an outline. In Information

Sciences, 172(1): 1-40.

Zadeh, L.A., 2006. From Search Engines to Question

Answering Systems – The Problems of World

Knowledge, Relevance, Deduction and Precisiation. In

Sanchez, E., ed., In Fuzzy Logic and the Semantic Web,

ed. Amsterdam: Elsevier, 163-210.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

360