Microsoft Word - brain_2_1.doc

Representing Mental Spaces and Dynamics of Natural Language Semantics

Ramin Golshaie
Department of Linguistics, Tarbiat Modares University, Tehran, Iran

ramin.golshaie@gmail.com

Abstract
Building systems with the robustness of human reasoning capabilities requires inspirations

from cognitive science. The primary objective of this study is to investigate the possibility of
representing some basic principles of cognitive semantics’ Mental Spaces Theory such as domain
construction, reality status of domains and their elements, and mental attitudes in a knowledge
representation framework for the purpose of developing cognitively plausible knowledge
representation systems. The model used as the basis of representation is the extended version of
conventional semantic networks, namely Multi-Layered Extended Semantic Networks (MultiNet).
The data used in this study have been selected from English expressions and have been represented
in MWR, MultiNet’s knowledge representation software. Results obtained from analysis of
represented data and their comparison to principles of mental spaces theory shows that theoretical
constructs of mental spaces theory such as domain construction, reality status of domains and their
elements, and mental attitudes can be formally represented in the MultiNet framework.

Keywords: Knowledge representation, mental spaces, semantic networks, conitive
semantics.

1. Introduction
In a daily conversation, we talk about a variety of objects, events, states of affairs, situations,

etc. Interesting above all is the fact that none of these conversational elements have to necessarily
be real. We may talk about yesterday’s car accident (a real event), admire Hercules Poirot (a
fictional detective), and even make a seemingly contradictory judgment that both Poirot is a
Belgian detective and in reality, Poirot is not Belgian are true! How could it be that an entity can
have a variable “reality status”? All of these facts can be accounted for by Mental Spaces Theory
(MST), a cognitive linguistic approach to meaning construction in natural language [7]. According
to Fauconnier [5, p. 351], “mental spaces are very partial assemblies constructed as we think and
talk for purpose of local understanding and action.” In other words, in the course of talking or
thinking we construct spaces in our mind all of which having their own internal structures and being
related to each other. Thus an entity can have a variable reality status depending on the mental
space to which it belongs. The mental space constructed by the sentence Poirot is a Belgian
detective is a non-real imaginary story space of which Poirot is an element. But when we say in
reality, Poirot is not Belgian the constructed space is reality space in which Poirot (the actor, not
the character) does not have the fictional nationality.

Figure 1. Mental space representation of “In the play, Mary is excited”

Knowledge representation, one of the main concerns of Artificial Intelligence (AI), is “the

study of how to put knowledge into a form that a computer can reason with” [18, p. 16]. Natural
language semantics as a rich source of knowledge is considered a major challenge to formalization
and representation. On the other hand, extracting semantic knowledge of natural language is a

R. Golshaie - Representing Mental Spaces and Dynamics of Natural Language Semantics

crucial task for natural language processing systems if they are to demonstrate a human-like
intelligence. In building meaning representation systems, various frameworks with different
orientations have been proposed: logic-based (Description Logics [1, 4]; Discourse Representation
Theory [11]), network-based (Semantic Networks [17]; Conceptual Dependency Theory [20];
Conceptual Structures [21]), and frame-based (Frame theory [14]) methods of meaning
representation are the most discussed ones in the literature.

One of the main problems in classical approaches to knowledge representation [15] is the
representation of beliefs about truth, and temporal aspects of knowledge. In response to these
difficulties [19], modal logics and the notion of possible worlds were introduced to account for
epistemic modalities inherent in the natural language. In the thesis of possible worlds [13] it is
strongly believed that there are lots of worlds and our world we inhabit is one of those possible
worlds.

Possible worlds, however, pose some metaphysical [2, p. 33] and referential [7] problems
not to be accounted for easily. In contrast, as Dancygier and Sweetser [3, p. 11] maintains, “mental
spaces represent a more general mechanism than possible worlds, referring not only to very partial
cognitive ‘world’ or ‘situation’ constructions as well as to more complete ones, but also to a variety
of non-world-like structures which can be connected and mapped onto other cognitive structures.”
Implementing mental spaces, as a cognitively motivated parallel to possible worlds, in a cognitive
computational framework would tackle the difficulties posed by logic-based methods. As Pereira
[16, p. 56] suggests, “From a symbolic AI perspective, a mental space could be represented as a
semantic network, graph in which we have nodes identifying concepts (corresponding to the
elements of a mental space) interconnected by relations. The definitions of mental space still allow
many other representations (e.g. cases in Cased-Based Reasoning, memes in Memetics or even the
activation pattern of a Neural Network in a given moment) but these would certainly demand more
complex computational treatment, especially with regard to the mapping.”

This paper aims at investigating the possibility of representing some principal notions of
mental spaces – such as domain construction, reality status of domains and their elements, and
mental attitudes – in a knowledge representation system. Our proposed framework, the Multilayered
Extended Semantic Networks (MultiNet), is a knowledge representation model developed by
Hermann Helbig [9, 8]. There are semantic analysis mechanisms built into MultiNet which are
hypothesized to show considerable overlap with basic principles of mental spaces. Comparison of
MultiNet representations and those of mental spaces theory shows that first steps can be taken to
make the outcomes of cognitive semantics research, specially mental spaces theory, realized and
pave the way for developing sophisticated and cognitively plausible knowledge representation and
reasoning systems.

2. Mental Spaces
Mental spaces are highly complex conceptual interconnected networks which are

constructed in the course of speaking or thinking. These conceptual networks or domains are
formed in the working memory and are expanded as the process of thinking or conceptualization
continues.

In the natural language, linguistic expressions serve as triggers in setting up mental spaces,
the levels at which meaning is also constructed. “These domains are not part of the language itself,
or its grammars; they are not hidden levels of linguistic representation, but language does not come
without them” [7].

Mental spaces, according to Fauconnier [6], are internally structured by frames and
cognitive models and externally are linked by “connectors” that relate mental spaces to one another.
New elements are added to spaces by linguistic and also non-linguistic expressions.

The followings are some of the linguistic devices used in constructing and linking mental
spaces [6, p. 40]:
• Space builders. A space builder is a grammatical expression that either opens a new space or

shifts focus to an existing spaces. Spaces builders take on a variety of grammatical forms,

BRAIN. Broad Research in Artificial Intelligence and Neuroscience
Volume 2, Issue 1, January 2011, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print)

such as prepositional phrases, adverbials, subject-verb complexes, etc.; for example, in 1929,
in that story, actually, in reality, in Susan’s opinion, Susan believes…, Max hopes…, if it
rains….

• Names and descriptions. Names (Max, Napoleon, etc.) and descriptions (the mailman, a
vicious snake, some boys who were tired, etc.) either set up new elements or point to existing
elements in the discourse construction.

In meaning construction there are also information about how elements of the spaces are

related. As an example we will consider mental space analysis of sentence (1):

(1) In the play, Mary is excited.

The space builder in example (1) is the phrase in the play, which sets up a mental space. In

Figure 1 the mental space is diagrammed by means of a circle and define a label (“PLAY”) to show
that the mental space represents the world inside the play. In this example, the name Mary
introduces a new element into the mental space which is labeled b. The expression excited in the
sentence, assigns a the property “EXCITED” to the element b. This information is included in the
‘box’ next to the mental space.

3. Meaning Representation in MultiNet
MultiNet is a knowledge representation framework (see [8]) developed by the Hermann

Helbig and his colleagues at IICS1 of University of Hagen. Its core design and functionality is based
on the notion of semantic networks. In this model concepts are represented with nodes in the
network. They are the smallest units of representation connected to one another by means of
explicitly defined relations and functions. A very brief overview of representational means in
MultiNet has been presented in the following subsections.

3.1. Concepts
One of the distinguishing features of MultiNet is its commitment to the Cognitive Adequacy

requirement. According this requirement [9], semantic representations and knowledge
representations should be centered around concepts. Concepts2 are represented by nodes in the
graphical representation of the network. Every node belongs to a specific sort defined by the
MultiNet’s ontology of sorts (Figure 2).

1 Intelligent Information and Communication Systems group
2 Following MultiNet convention, concepts are represented by the font concept in the text and multiword concepts are put in angled brackets:
<multi-word concepts>.

R. Golshaie - Representing Mental Spaces and Dynamics of Natural Language Semantics

Figure 1. The upper ontology of sorts in MultiNet (after Helbig [17])

There are seven attributes (layers) that characterize concept nodes in a multi-dimensional

space. A very brief definition for each, adopted from [9], has been given below:

• FACT: This attribute describes the “Facticity” of an entity, i.e. whether it is really existing
(value: real), not existing (value: nonreal), or hypothetically imagined (value: hypo).

• GENER: the “degree of generality” indicates whether a conceptual entity is generic (value:
ge) or specific (value: sp).

• QUANT: The intentional “quantification” represents the quantitative aspect of a conceptual
entity.

• REFER: This attribute specifies the “determination of reference”, i.e. whether there is s
determined object of reference (value: det) or not (value: indet).

• CARD: The “cardinality” as characterization of a multitude at preextensional level is the
counterpart of the attribute QUANT at the intensional level; it characterizes the number of
elements in a set.

• ETYPE: This attribute characterizes the “type of extensionality” of an entity with values: nil –
no extension, 0 – individual which is no set, 1 – entity with a set of elements from type
[ETYPE 0] as extension, 2 – entity with a set of elements from type [ETYPE 1] as extension.

• VARIA: The “variability” describes whether an object is conceptually varying (value: var) – a
so-called parametrized object – or not (value: con).

BRAIN. Broad Research in Artificial Intelligence and Neuroscience
Volume 2, Issue 1, January 2011, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print)

Concepts need to be connected to one another in order to make realization of situations
possible. These connections are established by a set of built-in relations and functions.

3.2. Relations
MultiNet has a total number of 89 relations for maintaining the conceptual dependencies

between nodes. 16 relations are C-role (Cognitive Role) and they are those relations that describe
the relationships between the main participants of a situation. As an example consider the sentence
Peter finished the discussion., semantic representation of which is given in Figure 3.

Figure 2. Semantic representation of the sentence Peter finished the discussion in MultiNet after Helbig [8, p. 447].

Semantic frame of the verb finish requires two C-roles: AGT (agent) and AFF (affected). The relation SUBS indicates
that the situational concepts finish and discussion are subordinate to their relevant generic concepts.

In Figure 3, semantic frame of the concept Finish realized in the form of the verb finish

requires two C-roles: An agent represented by the relation AGT, and an affected entity represented
by the relation AFF. Here agent is Peter and the affected entity is an abstract object ([SORT = ad]
means the concept is a dynamic abstraction). The following hierarchy in MultiNet ontology leads to
the sort [ad]:

Entities [ent] > objects [o] > abstract objects [ab] > abstractions from situations [abs] > dynamic
abstraction [ad].

The relation SUBS indicates that our concept in this particular situation is subordinated to
its relevant generic concept.

Now the basic concepts of MST and MultiNet framework have been introduced and we are
ready to see if MultiNet is capable of formally representing or simulating basic mechanisms of
mental spaces.

4. Mental Spaces and Their Computer Modelling
As mentioned previously, MultiNet is a knowledge representation framework for

representing semantic content of natural language expressions. It is not a framework for
implementing MST. This section will discuss the MultiNet’s potentiality in capturing and
representing MST’s basic principles in the construction of domains, their truth, and reality status of
their elements.

4.1. Method
In the process of building mental spaces, some natural language expressions (space builders)

play the primary role in setting up spaces. In order to make a systematic analysis for the purpose of
this paper, space builders (taken from English expressions) were classified into three classes
considering theirs grammatical form [6]: Prepositional phrase space builders (PPSB) like in that
story…, subject-verb space builders (SVSB) like he believes…, and conditional space builders
(CNSB) like if it rains….

In MultiNet, the representational means corresponding to space builders – which were
formally classified into three categories – were searched for in MultiNet relations based on
definitions provided for them [8]. After a careful comparison, the three categorized space builders

R. Golshaie - Representing Mental Spaces and Dynamics of Natural Language Semantics

were found to be best comparable with MultiNet relations presented in Table 1. To test this
resemblance, some simple English sentences containing space builders of the classified three types
were constructed and their MST and MultiNet representations were compared.

Table 1. Mental spaces builders and their corresponding relations in MultiNet.

4.2. Types of Space Builders
In this part, the aforementioned three types of space builders will be elaborated with an

example.

4.2.1. Prepositional Phrase Space Builders
Prepositional Phrase Space Builders’ (PPSBs) main role is to explicitly restrict the

description of a situation to a particular context. As a simple example, consider the sentence in (2):

(2) In the film, John is riding a unicorn.

Figure 4. The mental spaces set up by the first interpretation of the sentence in the film, John is riding a unicorn.

In the sentence (2), the prepositional phrase in the film is a space builder; it constructs a fictional
space relative to the reality space. Here, sentence (2) can be interpreted in two different ways:

• First interpretation: The reality space (let’s call it R) contains an element a associated with
the proper name John. The noun phrase a unicorn introduces an element b׳ to the film space
(call it F). I is the connector linking a in the space B to a׳ in the space F (Figure 4). Since the
elements of both mental spaces are co-referential, this connector is an identity connector.
The rectangles represent the internal structure of the spaces next to them. The dashed line
indicates that the space F is set up in relation to R and that it is subordinate to R in
discourse.

3 Scope of the SVSBs (subject-verb complexes) in this study is limited to verbs characterizing mental processes.

Mental space builders MultiNet relations

PPSB CTXT (Restricting context)

SVSB3 MCONT (Mental content)

CNSB COND (Conditional relation)

BRAIN. Broad Research in Artificial Intelligence and Neuroscience
Volume 2, Issue 1, January 2011, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print)

• Second interpretation: The proper name John exists only in film space without having a
counterpart in base space. Therefore, the second interpretation of the sentence is: John is a
film character who is riding a unicorn in the film (Figure 5).

Figure 5. The mental spaces set up by the second interpretation of the sentence in the film, John is riding a unicorn.

4.2.2. Subject-Verb Space Builders
Subject-Verb Space Builders (SVSBs) are the second type of our classified space builders.

The verbal position of SVSBs can be filled with a variety of verbs among which are mental or
psychological verbs such as think, believe, suspect, etc. representing mental attitudes. For example,
consider the sentence in (3):

(3) Mary thinks that John smokes.

The proper nouns Mary and John setup a base space (B). By the help of background
knowledge and activated frames we know that they are names of female and male humans. Not
having access to the previous discourse, we also consider their existence presupposed. The SVSB
Mary thinks that sets up a belief space (L) relative to space B (Figure 6). The identity connector
maintains the referential link between elements a and a׳ both referring to the same person.

Figure 6. The mental spaces set up by the sentence Mary thinks that John smokes.

4.2.3. Conditional Space Builders
The third type of space builders, according to our classification, are those involving

Conditional Space Builders (CNSBs) like if it rains…, if something goes wrong…. if element sets up

R. Golshaie - Representing Mental Spaces and Dynamics of Natural Language Semantics

a hypothetical space in which the situation described may or may not come true. Consider the
example in (4):

(4) If John buys the car, he will drive to Berlin.

In the sentence (4) there are three proper names which constitute elements of the base space
or reality space (R). The CNSB if sets up the hypothetical space (H) with elements identical to those
of reality space (Figure 7). Every space’s internal structure is presented in the boxes next to them.

In the next part, we will see how MultiNet’s built-in meaning representation mechanisms are
capable of representing the basic principles of mental space building outlined above.

Figure 7. The mental spaces set up by the sentence If John buys the car, he will drive to Berlin.

4.3. Analysis of MultiNet Representation
This part will present MultiNet representations of different mental spaces discussed

previously. MultiNet has a graphical representation environment called MWR (abbreviated from the
German “MultiNet Wissens Repräsentation” which is translated into English as MultiNet
Knowledge Representation) which has a powerful and user-friendly facilities for making networks
out of concept nodes and relational dependencies.

BRAIN. Broad Research in Artificial Intelligence and Neuroscience
Volume 2, Issue 1, January 2011, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print)

Figure 8. Comparison of MultiNet and MST representations.

Top: Representing first interpretation of the sentence in the film, John is riding a unicorn in MultiNet. Bottom:
Corresponding mental space representation (color markings added for illustration).

4.3.1. CTXT: Restricting Context
According to Helbig [8, p. 40], “[The] relation [CTXT] restricts the whole conceptual

capsule and its content to ‘a certain world’ (world view) or a certain context.” Here the phrase ‘a
certain world’ can be interpreted as equivalent to a new space in MST which is set up relevant to its
base space. For clarification, consider the sentence in (2) given below as (5):

(5) In the film, John is riding a unicorn.

R. Golshaie - Representing Mental Spaces and Dynamics of Natural Language Semantics

Figure 9. Comparison of MultiNet and MST representations.

Top: Representing first interpretation of the sentence in the film, John is riding a unicorn in MultiNet. Bottom:
Corresponding mental space representation (color markings added for illustration).

We saw that the sentence (5) had two interpretations. Their corresponding MultiNet

representations should also reflect these two readings.
According to the above definition, the PPSB in the film can roughly be taken as equivalent

to CTXT relation in MultiNet. The relation CTXT restricts the situation to the film space in which
the elements unicorn and John have different reality status. Unicorn is a fictional animal, so it is
characterized by the attribute-value [FACT = nonreal]. The first interpretation of the sentence in
mental spaces, characterized John as an element of reality space. In MultiNet this can be achieved
by specifying the attribute-value [FACT = real] for the concept John. Facticity attribute-value of the
attsituation sv1 is also characterized as [FACT = nonreal]. So the first reading of the sentence (5) is
represented in MultiNet with John having [FACT = real] attribute-value (belonging to reality space)
and other entities having [FACT = nonreal] attribute-value. Figure 8 shows the representation of the
first reading of the sentence (5) in MWR. The MultiNet is shown at top of the figure, and the

BRAIN. Broad Research in Artificial Intelligence and Neuroscience
Volume 2, Issue 1, January 2011, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print)

corresponding mental space representation is shown at the bottom side. Blue circles mark
hypothetical or nonreal objects and situations and red circles mark real objects and situations.
Facticity value for John is real, for unicorn it is nonreal, and for the process of riding, marked with
blue broken circle, it is non-real as well.

Second interpretation of the sentence (5) can be represented with changing John’s Facticity
attribute-value from [FACT = real] to [FACT = nonreal] making the whole situation and its
elements non-real (Figure 9).

4.3.2. MCONT: Mental Content
Helbig [18, p. 507] defines the expression (s MCONT c) as “a specification of the

informational or mental content c of a mental or informational process s… By default, the second
argument c is assumed to be a hypothetical object or situation.”

MCONT relation properties are roughly equivalent to those of subject-verb complexes
(SVSBs) of MST. MCONT relation can generally be treated as capturing the idea behind what
philosophers call propositional attitudes in representational theories of mind [12] or opaque
contexts [10] in semantics. For example, consider the sentence in (3) rewritten below as (6) and
semantic representation of which is given in Figure 10:

(6) Mary thinks that John smokes.

Sv1 describes a situation in which we have the concept represented by the verb think that
requires a “mental experiencer” (MEXP). In the frame of the verb think there is also another
argument specifying the content of thinking process (MCONT) which according to definition is
hypothetical ([FACT = hypo]) by default. Sv2 centers around the concept smoke which is realized
by the verb smoke and requires one agent (AGT) argument. Here we suppose that the verb smoke is
intransitive so the concept smoke needs just one argument. In sv2 the Facticity of the concept John
is set to [real] implying that john exists in reality (base) space and it is just the whole situation sv2
that is hypothetical.

In representing the sentence Mary thinks that John smokes, Mary and John (marked with red
circles) and also the process of thinking (marked with red broken circle) are elements of reality
space, and process of smoking (marked with blue broken circle) belongs to Mary’s belief space.

R. Golshaie - Representing Mental Spaces and Dynamics of Natural Language Semantics

Figure 10. Comparison of MultiNet and MST representations.

Top: Semantic representation of the sentence Mary thinks that John smokes in MWR. Bottom: Corresponding mental
space representation (color markings added for illustration).

4.3.3. COND: Conditional Relation
If situation sv1 is said to be a sufficient condition for the situation sv2, the relation COND

can be established between the two situations: sv1 COND sv2. This means that sv1 is a trigger for
sv2 while both situations are hypothetical (FACT = hypo). In MST’s terms we can say both of the
situations are set up in a hypothetical space. Consider example (4) given below for convenience as
(7):

(7) If John buys the car, he will drive to Berlin.

In the sentence (7) there are two clauses describing two hypothetical situations. These two
situations are connected by the relation COND in MultiNet representation (Figure 11).

In the mental spaces constructed by the sentence (7), we saw that the nominals John, the car,
Berlin all were set up in the base space which are marked with red circles. Similarly in MultiNet’s

BRAIN. Broad Research in Artificial Intelligence and Neuroscience
Volume 2, Issue 1, January 2011, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print)

representation these objects’ Facticity value is set to real: [FACT = real]. This means if we take the
reality space as our base space, entities having the attribute-value [FACT = real] are elements of
base space. On the other side, entities or situations having the attribute-value [FACT = hypo] belong
to the hypothetical space: sv1 and sv2 in our example are hypothetical situations marked with blue
broken circles to indicate hypothetical quality of their representative processes.

Figure 11. Comparison of MultiNet and MST representations.
Top: Semantic representation of the sentence if John buys the car, he will drive to Berlin in MWR. Bottom:

Corresponding mental space representation (color markings added for illustration).

R. Golshaie - Representing Mental Spaces and Dynamics of Natural Language Semantics

5. Concluding Remarks
Mental Spaces is a theory about human knowledge representation and processing. It has

proved to be successful in accounting for some cognitive phenomena such as constructing
conceptual domains, an entity’s variable reality status, and mental attitudes. These phenomena pose
serious problems in building automatic knowledge processing and reasoning systems.

In this paper, I tried to establish a connection between meaning representation mechanisms
of mental spaces and those of MultiNet, a knowledge representation framework. In building mental
spaces, some linguistic expressions, called space builders, act as a trigger to construct mental
domains. Some major space builders were formally categorized into three classes: prepositional
phrase space builders (PPSBs), subject-verb complexes space builders (SVSBs), and conditional
space builders (CNSBs). The correspondence between MST space builders and MultiNet relations
can be stated as following points:

In building mental spaces, some linguistic expressions, called space builders, act as a trigger
to construct mental domains. Some major space builders were formally categorized into three
classes: prepositional phrase space builders (PPSBs), subject-verb complexes space builders
(SVSBs), and conditional space builders (CNSBs). The correspondence between MST space
builders and MultiNet relations can be stated as following points:

1. PPSBs in MST have their counterpart as CTXT relation in MultiNet. CTXT relation restricts

the validity of a situation to a specific context or world.
2. SVSBs (of the type mental processes) in MST have their counterpart as MCONT relation in

MultiNet. MCONT’s second argument is hypothetical by default implying the situation
described by the second argument is constructed in a hypothetical space.

3. CNSBs in MST have their counterpart as COND relation in MultiNet. Both arguments of
COND relation are hypothetical by default. Thus the hypothetical situations are set up in a
hypothetical space.

Reminding the paper’s objectives, this study aimed to investigate formalization of MST’s
representational concepts such as domain construction, reality status of domains and containing
elements, and mental attitudes in a knowledge representation framework. To this end, MultiNet
framework provides us with the representational means summarized in table 2.

Table 2. MST concepts and corresponding MultiNet representational mechanisms.

Scope of this study was limited to basic constructions and principles. There are still many
semantic phenomena not mentioned here, but studied and handled in the MST framework:
presuppositions, tense and mood, counterfactuals, etc., are just some of these. This study can be
further extended to include the untouched domains.

MultiNet

Relations Attributes

Domain construction CTXT, MCONT, COND FACT

Reality status of domains
and elements --------- FACT MST

Mental attitudes MCONT ---------

BRAIN. Broad Research in Artificial Intelligence and Neuroscience
Volume 2, Issue 1, January 2011, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print)

6. Acknowlwdgment
I am grateful to Ingo Gloeckner (IICS, University of Hagen) who kindly provided me with

MWR software.

7. References

[1] Baader, F., I. Horrocks, and U. Sattler (2008), Description logics. In: F. van Harmelen, V.
Lifschitz, B. Prter (Eds.), Handbook of Knowledge Representation, Elsevier, pp. 136-169.
[2] Croft, W., and D. A. Cruse (2004), Cognitive Linguistics, Cambridge University Press,
Cambridge.
[3] Dancygier, B., and E. Sweetser (2005), Mental Spaces in Grammar, Cambridge University
Press, Cambridge, 2005.
[4] Donini, F., M. Lenzerini, D. M. Nardi, and A. Schaerf (1996), Reasoning in description logics.
In: G. Brewka (Ed.), Principles of Knowledge Representation: Studies in Logic, Language and
Information, CSLI Publications, pp. 193-238.
[5] Fauconnier, G. (2007), Mental spaces. In: D. Geeraerts, H. Cuyckens (Eds.), Oxford Handbook
of Cognitive Linguistics, Oxford University Press, Oxford, 2007, pp. 351-376.
[6] Fauconnier, G. (1997), Mappings in Thought and Language, Cambridge University Press,
Cambridge.
[7] Fauconnier, G. (1994), Mental Spaces: Aspects of Meaning Construction in Natural Language,
Cambridge University Press, Cambridge.
[8] Helbig, H. (2006), Knowledge Representation and Semantics of Natural Language, Springer,
Berlin.
[9] Helbig, H. (2002), The use of multilayered extended semantic networks for meaning
representation. In: G. Katz, S. Reinhard, P. Reuter (Eds.), Sinn & Bedeutung VI: Proceedings of the
Sixth Annual Meeting of the Geselleschaft für Semantik, University of Osnabrück.
[10] Hurford, J. R., B. Heasley, and M. B. Smith (2007), Semantics, 2nd Ed., Cambridge University
Press, Cambridge.
[11] Kamp, H., and U. Reyle (1993), From discourse to logic: introduction to model-theoretic
semantics of natural language, formal logic and discourse representation theory, Number 42 in
Studies in Linguistics and Philosophy, Kluwer Academic Publishers, Dordrecht, The Netherlands.
[12] Kolak, D., W. Hirstein, P. Mandik, and J. Waskan (2006), Cognitive Science: An Introduction
to Mind and Brain, Routledge, New York.
[13] Lewis, D. (1986), On the Plurality of Worlds, Blackwell, Oxford.
[14] Minsky, M. (1975), A framework for representing knowledge. In: P. H. Winston (Ed.), The
Psychology of Computer Vision, McGraw-Hill, New-York, pp. 211-277.
[15] Partridge, D. (1996), Representation of knowledge. In: M. A. Boden (Ed.), Artificial
Intelligence, Academic Press, California, pp. 55-85.
[16] Pereira, F. C. (2007), Creativity and Artificial Intelligence: A Conceptual Blending Approach,
Mouton de Gruyter, Berlin.
[17] Quillian, M. R. (1968), Semantic memory. In: M. Minsky (Ed.), Information Processing MIT
Press, Cambridge Massachusetts, pp. 227-270.
[18] Russel, S. J. and P. Norvig (2003), Artificial Intelligence: A Modern Approach, 2nd Ed,
Prentice-Hall, New Jersey.
[19] Saeed, J. (1997), Semantics, Blackwell, UK.
[20] Schank, R. C. (1972), Conceptual dependency: A theory of natural language understanding,
Cognitive Psychology, 4, 532-631.
[21] Sowa, J. F. (1984), Conceptual Structures: Information Processing in Mind and Machine,
Addison-Wesley, Massachusetts.