Cognitive Modeling in Computational Rhetoric: Litotes, Containment

and the Unexcluded Middle

Jelena Mitrovi

1 a

, Cliff O’Reilly

2 b

, Randy Allen Harris

3 c

and Michael Granitzer

1 d

Faculty of Computer Science and Mathematics, University of Passau, Germany

Birkbeck College, University of London, U.K.

Department of English Language and Literature, University of Waterloo, Canada, U.S.A.

Keywords:

Litotes, Computational Rhetoric, Cognitive Modeling, Ontology, Image Schema, Unexcluded Middle.

Abstract:

The focus of our study is the rhetorical ﬁgure litotes and its cognitive modeling. This ﬁgure is formed by a

contrary term that emphatically accentuates a positive, e.g. He is not exactly an idiot (said of Albert Einstein).

Lawrence Horn’s illumination of the Law of Excluded Middle and its relationship to litotes by creating an

Unexcluded Middle is central to our ideas and we correlate this to Image Schema theories developed by Mark

Johnson, George Lakoff and Rafael N

nez – speciﬁcally the schema of CONTAINMENT. The distinction be-

tween contrary and contradictory opposition is described. We extend the assessment of the Excluded Middle

from the perspective of Image Schema theory into the realm of the Unexcluded Middle and draw a represen-

tation of the layout of containers and analogous concept-activation. Lastly, we create and present an OWL

ontology and publish it online.

1 INTRODUCTION

We argue for the integration of rhetorical studies,

cognitive science, computational modeling, and Nat-

ural language processing (NLP), by way of an ex-

tended example, the unjustly neglected rhetorical ﬁg-

ure, litotes.

Litotes is a ﬁgure in which we say less but mean

more (“minus dicimus et plus signiﬁcamus” (Miguel,

1990)) and can be placed in the Pragmatic context un-

der Implicature – the difference between what is said

by a speaker and what is intended

. Litotes is as ubiq-

uitous as any of the more famous rhetorical ﬁgures

such as metaphor and irony, and yet not well studied

from a computational perspective. Despite this ne-

glect, we hold that research on its function, role and

importance promises headway in the rich new ﬁeld

of ﬁgure analysis in Computational Rhetoric, an area

of study which brings together computer scientists,

rhetoricians, psychologists, philologists, literary crit-

ics and information scientists working on the detec-

https://orcid.org/0000-0003-3220-8749

https://orcid.org/0000-0001-8002-5336

https://orcid.org/0000-0002-9324-1879

https://orcid.org/0000-0003-3566-5507

For a deﬁning overview of rhetorical ﬁgures including

litotes see (Harris and Di Marco, 2017)

tion and understanding of rhetorical ﬁgures, their as-

sociated cognitive machinery and computational ap-

plicability.

In this paper we connect the realm of rhetorical

studies to the domain of Image Schema theory. As

Dewell has said: “...language plays a much greater

role in the development of image schemas than is of-

ten assumed, contributing not only to cross-linguistic

variation but also to some universal similarities in

the structure of image schemas” (Dewell, 2005). We

are especially concerned with the universal schema

known as containment, which gives a valuable pur-

chase on litotes.

The Law of the Excluded Middle is a primary el-

ement of logical thought and a variant extension of

this law – named the Unexcluded Middle – effectively

bridges litotic understatement and cognitive concept-

activation, i.e. the neuronal activity resulting in con-

scious or subconscious understanding of a concept.

In section 2 we offer some background on litotes,

a more detailed analysis of the Unexcluded Middle,

the Image Schema CONTAINMENT and the onto-

logical approach to modeling litotes. Section 3 gives

an analysis of how these concepts are intertwined in

litotes while in section 4 we provide an OWL ontol-

ogy of litotes. Conclusions and Future work consid-

erations are given in the last two sections of the paper.

806

Mitrovi

c, J., O’Reilly, C., Harris, R. and Granitzer, M.

Cognitive Modeling in Computational Rhetoric: Litotes, Containment and the Unexcluded Middle.

DOI: 10.5220/0009104908060813

In Proceedings of the 12th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2020) - Volume 2, pages 806-813

ISBN: 978-989-758-395-7; ISSN: 2184-433X

2 BACKGROUND

2.1 Litotes

Litotes is a rhetorical ﬁgure in which a state-

ment is made emphatic by denying a con-

trary. A straightforward example from Henry

Peacham’s The Garden of Eloquence (1593) is

“he is not the wisest man in the world” which has a

meaning similar to “he is a fool” (Lanham, 1991).

It is a form of understatement, also known as

Deminutio to Latin pre-Christian scholars, but the

term comes to us from the Greek meaning “smooth”

or “plain”: λιτοτης. An example of litotes appears in

the early handbook erroneously attributed to Cicero,

the Rhetoric ad Herennium (Cicero, 1954):

“his father left him a patrimony that was – I

do not wish to exaggerate – not the smallest”

Cicero expresses the point of mitigating a posi-

tive but nevertheless getting across the message in-

tended, i.e. that the legacy in question was large in-

deed. There is, in short, some level of irony at work

with litotes, and interpretation requires a reconstruc-

tion of the speaker’s intention. The impact of litotes

derives from speaking plainly about a subject while,

at the same time, contradicting a negative term to

enhance the positive. Its usefulness as a means for

understatement, modesty and insult leads to signiﬁ-

cant usage and an almost ubiquitous position in lan-

guage and culture (Shovel, 2015). Its prominence has

brought a few interesting studies in more recent times

(Hermann et al., 2013; Yuan, 2017)

Horn examines the ﬁgure in great detail (Horn,

2017) and differentiates between two important con-

cepts - contrariety and contradiction. A contradictory

opposition is binary, e.g. black versus white or green

versus not green, whereas a contrary opposition can

allow room for things to lie in-between, e.g. truthful

versus untruthful or comfortable versus uncomfort-

able. Nuance, context and vagueness are important

factors that determine the impact of the differences in

contrary oppositions.

Litotes aligns both with the principles of logic (the

negation of a negative is a positive) and the prin-

ciples of style (no repetition is null). Repetitions

work chieﬂy in an additive way, but the main is ad-

ditive, familiar in the iconicity principle of quantity –

more language corresponds to more conceptual mat-

ter (Givon, 1991). That is where the emphasis comes

in. The two negatives in this case are not just positive,

they are emphatically positive. Denotatively, “not the

smallest” means that there is at least one other quan-

tity that is smaller, the not cancelling out the superla-

tive. Rhetorically, it means something more like “the

biggest”, or at least “really big”. Sometimes, how-

ever, repeating the negative has less emphatic sense.

For instance, compare the two sentences, “Lupita Ny-

ong’o is not unattractive” (with a negator and a negat-

ing preﬁx) and “Lupita Nyong’o is not ugly” (with a

negator and a negatively valenced adjective). The ﬁrst

one is mildly complimentary about her appearance,

the second more emphatic. And with the following

example, which pushes the disavowal of the adjective

to an extreme, the compliment approaches perfection:

“Lupita Nyong’o is the furthest thing from ugly.”

Computational modeling of litotes has to take into

consideration its many characteristics, negation being

the most prominent. Negation has been treated differ-

ently in Linguistics and Rhetoric, as well as in Prag-

matics, where it has a strong effect (Taboada et al.,

2017). Likewise, a rhetorical ﬁgure such as Antithesis

has been seen as very effective and even persuasive,

precisely due to the presence of negation, as well as

opposition. We see litotes as a ﬁgure that can also

play a persuasive role in arguments, therefore its de-

tection should be included in argument mining sys-

tems (Mitrovi

c et al., 2017).

2.2 Unexcluded Middle

In logic, the Law of Excluded Middle, dating at

least to Aristotle’s discussion of the principle of non-

contradiction in On Interpretation (Aristotle, 1938),

states that for any proposition, either that proposition

is true or its negation is true. It is the third of the three

classic laws of thought. This, now (in)famous quote

from a former United States’ president illustrates us-

age of the excluded middle in modern practice :

(1) Either you are with us or you are with the

terrorists (George W. Bush)

Using propositional logic notation to reﬂect the logic

of the terms and arguments involved we can say that

P ∨ Q becomes P ∨ ¬P (P represents being “with us”

and Q represents being “with the terrorists”). By im-

plication anyone not with “us” must be with “the ter-

rorists”. We illustrate this in Figure 1 – we transi-

tion from a three member set {us,terrorists,neither us

nor terrorists} to a two-member set {us,terrorists} and

the middle is effectively excluded from being (“po-

lar contraries p and q become mutually exhaustive as

well as mutually inconsistent” (Horn, 2017)).

Figure 1: Transition to Excluded Middle.

Cognitive Modeling in Computational Rhetoric: Litotes, Containment and the Unexcluded Middle

807

There is a further twist to this pattern called the

Unexcluded Middle. This aspect of the litotic form

can be seen in the following example:

(2) I’m not happy but I’m not unhappy about

it. (The History Boys, Alan Bennett)

The speaker is neither happy nor unhappy which vio-

lates the law of non-contradiction, but rhetorically the

meaning is that happiness is irrelevant to the situa-

tion. Pragmatically, we may also say that (despite the

binary nature of the adjectives) happiness can form

a contrary opposition, not contradictory one, so that

the speaker can be seen to articulate a midpoint on a

continuum, neither positive (happy) nor negative (un-

happy), the negator and the negative afﬁx cancelling

each other out. Figure 2 shows a representation of the

Unexcluded Middle and the creation of a new state

that is in-between happy and not happy. We could

graphically represent this in a number of ways, but

choose this three-member set for simplicity.

Figure 2: Representation of the Unexcluded Middle.

Reinhard Blutner describes this phenomenon as

“double negation” and draws it slightly differently

(Figure 3) (Blutner et al., 2004). Blutner also reﬂects

that “it’s an interesting exercise to introduce more

than three states of happiness...More importantly, in

the context of litotes, it seems necessary to account

for the effect of gradient acceptability and continuous

scales” (Blutner et al., 2004).

Horn states that “To depict individuals (or their

marriages and routines) as ’not unhappy’ yet – by

implication – not (exactly) happy either is to tac-

itly invoke a scale <not unhappy, happy> in accor-

dance with which a weaker property is afﬁrmed or

conceded but a stronger alternative implicitly denied”

(Horn, 2017). The key term for us here is “by im-

plication” since it is the mental process of blending

knowledge about what happiness in marriage is like

which can give rise to opposition. What is implied by

understanding of words, contexts, stereotypes, Scripts

(Schank and Abelson, 1975), Frames (Fillmore et al.,

1982) etc. affects what happens in the balancing of

terms.

Figure 3: When two negatives don’t make a positive, repro-

duced from (Blutner et al., 2004).

Just as regarding afﬁxal negation (the

“notorious not un- construction” (Horn, 2017))

such as “not unhappy”, we can include a similar

description of a term that creates a contradictory

opposition. In the example phrase, below, by Roger

Penrose, the Unexcluded Middle is created between

the possible and impossible.

(3) Quantum entanglement is a very intriguing

issue, but it is not impossible. (Roger Penrose)

Possibility should not be able to be deﬁned in this way

(Frege’s assertion that “clothing a thought in dou-

ble negation does not alter its truth value” (Frege

and Beaney, 1997) should hold true for proposi-

tional logic), however this example shows another

Unexcluded Middle where understanding the sub-

ject (Quantum Entanglement) is in some way nei-

ther possible nor impossible. This case could be for

the purposes of understatement or simply a play on

the concept of impossibility given that the perceived

difﬁculty of understanding Quantum Theory is well

known. Horn’s analysis deﬁnes a symbolic represen-

tation for this phenomenon:

1. ¬(

[

adj

X])

2. ¬(¬[

adj

X])

(

 represents contrary opposition, ¬ represents con-

tradictory opposition and [

adj

X] represents an adjec-

tival term.)

2.3 Image Schemata

Image Schemata are “structures by which we are able

to have coherent experience that we can comprehend”

(Johnson, 1987), but “they can be seen as the con-

ceptual building blocks for metaphoric and abstract

thought” (Hedblom et al., 2015). Image Schemata

provide a framework which we can use to describe

complex structures of understanding and cognition in

many domains. We focus herein on the linguistic,

but also take into account that the underlying mecha-

nisms are neurological, develop from infancy and are

sensori-motor in origin (Cuccio and Gallese, 2018).

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

808

Figure 4: Venn-style diagram representing CONTAINER

objects.

2.3.1 The Schema CONTAINMENT

There are many recognized Image Schemata, one fun-

damental example being CONTAINMENT. Johnson

discusses the structure of this schema. Based on ex-

perience of the world (grounded or embodied), a re-

curring organization of structure, spatially bounded,

a three-dimensional container is experientially salient

and can be thought of as an “in-out orientation”

(Johnson, 1987).

Lakoff and N

nez approach core concepts in

mathematics from the direction of Image Schema the-

ory (Lakoff and N

nez, 2000). Of particular interest

to us is the treatment of CONTAINMENT mapped to

fundamental inferential laws of logic and especially

the Law of Excluded Middle. Also the link estab-

lished between visual processing in the brain (a root

for Image Schema theory) and concept-activation ap-

plied to non-visual stimuli, gives us a basis for the

arguments in this paper and therefore a thesis that the

underlying mechanisms of language understanding –

especially the complex and rich domain of rhetoric

– are predicated on the same brain structures (or at

least connected to them) that give rise to visual pro-

cessing and conceptualizations; “...it makes neurolog-

ical sense that structures in the visual system can be

used for conceptual purposes” (Lakoff and N

nez,

2000). As argued further in their work, the capacity

for perceiving the world in terms of contained struc-

tures gives rise to a sort of Folk Boolean Logic which

can be associated to a Venn diagram, such as in Fig-

ure 4. Container A represents a concept. Object B is

within the container and therefore adheres to the con-

cept of A.

Each region in the diagram is related to a con-

cept that maps to a Container element (either in or

out of the container). Lakoff and N

nez develop a

Law of Container Schemas which evolves from the

mapping of structural constraints (from the physical

process of containment and exclusion) to inferential

structures which are conceptual in nature, but fol-

low analogous “rules”, e.g. an object inside a con-

tainer maps to a member of the category represented

by the container. This gives rise to the Conceptual

Metaphor Categories Are Containers. Table 1 shows

the progression through logic from the CONTAINER

schema to a symbolic representation of the Law of

Excluded Middle.

2.4 Ontological Approach

Ontologies are ideal for representing ideas as com-

plex as rhetorical ﬁgures, which have speciﬁc prop-

erties and are often interrelated (Harris et al., 2017).

The ﬁgure of Anadiplosis, for example, has the prop-

erty of repeating lexemes across clause boundaries

and litotes, as we have seen, is centred on spe-

ciﬁc phrases (lexemes) in well-deﬁned orientations

(double-negation). Furthermore, rhetorical ﬁgures

have a tendency to co-occur which makes linking con-

cepts within and across distinct ontology models a

useful tool for extending understanding and also for

inferential knowledge discovery.

Our ontology of litotes is modeled and developed

in OWL

which enables us to utilize the power of log-

ical inference and validation with tools such as Pro-

tege

. It also enables signiﬁcant re-use within the Se-

mantic Web movement since, by publishing the on-

tology ﬁles online, we enable others to beneﬁt from

our work. We create links via OWL properties to on-

tologies already part of the Semantic Web and this po-

tentially allows inference and computation across the

combined set of data we hope to create in the near

future.

There are numerous recent works that develop

an ontological perspective on image schemata (Kim-

mel, 2005; Kuhn, 2007; Hedblom et al., 2014; Hed-

blom et al., 2018), however, to our knowledge, few

OWL ontologies exist that describe the cognitive as-

pects of rhetorical ﬁgure sub-structure (a number of

taxonomically-focussed ontologies have been devel-

oped however, e.g. the Ontology of Rhetorical ﬁg-

ures for Serbian (Mladenovi

c and Mitrovi

c, 2013)

and the Ontology focusing on the ﬁgure Antimetabole

(O’Reilly and Harris, 2017).

3 ANALYSIS

Similar to Bennett & Cialone’s analysis of geomet-

rical constraints through their “Eight kinds of sur-

rounding” (Bennett and Cialone, 2014) we explore

the Venn-style representation of the container/concept

analogy. Lakoff and N

nez represent the Container

Schema logic for Excluded Middle as a container with

https://www.w3.org/OWL/

https://protege.stanford.edu/

Cognitive Modeling in Computational Rhetoric: Litotes, Containment and the Unexcluded Middle

809

Table 1: Logic and Law progression for Excluded Middle (from (Lakoff and N

nez, 2000).

Logic Law

Embodied Container Schemata Every object X is either in Container schema A or out of

Container schema A

Mapped classes Every entity X is either in category A or out of category A

Propositions mapped from classes Every proposition P is either true or not true

Symbolic form mapped from propositions P ∨ ¬P

an entity X either in or out of the container A – Figure

Figure 5: Container Schema representation for Excluded

Middle, reproduced from (Lakoff and N

nez, 2000).

To follow our previous example of an utterance cred-

ited to the former US President Bush, we might say

that the container A represents the category terrorist

and for entity X to be positioned in the container is to

be a terrorist and outside the container is to belong to

us as presented in the infamous quote given in Section

2.2 of this paper.

In considering the Unexcluded Middle we extend

this analogy and diagramming method. Figure 6

shows a view of container and entity orientation in re-

spect to the Unexcluded Middle. To follow the exam-

ple from Penrose again, an opposition is created be-

tween possible and impossible, but because the “se-

mantic contradictory is also a virtual (or coerced)

contrary” (Horn, 2017) we have to consider three cat-

egories and therefore three containers.

Figure 6: Container Schema representation #1 for Unex-

cluded Middle.

The process of double negation similar to those de-

scribed herein leads to concept-activation (again, we

are not referring exclusively to the creation of a novel

concept representation) and a conceptual leap that re-

sults in added complexity and ambiguity. Container

A could be possible, container B impossible and con-

tainer C could represent not impossible (the so-called

Unexcluded Middle). The entity X is either in con-

tainer A, B or C.

Figure 7: Container Schema representation #2 for Unex-

cluded Middle.

We contend that in reality it is plausible that some

form of all three concepts (containers) are activated

especially where concepts are not completely con-

trary or if there is a particularly large contextual over-

lap. We show this alternative representation in Fig-

ure 7. The key difference between Figures 6 and 7

is that the containers are merged or in contact (John-

son lists MERGING and CONTACT as separate im-

age schemata (Johnson, 1987) which indicates some

overlap or connection between the concepts.

4 ONTOLOGY

We take the previous insights gained from coalesc-

ing the image schema of CONTAINMENT, our un-

derstanding of linguistic entities, and the concept-

activation process we have termed the Unexcluded

Middle and develop an OWL ontology representation.

We draw a graph showing the main concepts and re-

lations and display it in the Figure 8.

We describe the linguistic entity Lexeme which

may be proximal-to any other lexeme and each of

which may evoke a Sememe. The sememe being a

semantic unit and lexeme being a lexical unit – these

terms reﬂect linguistic standards. A sememe can be

situated-in either an Exterior or Interior of a Con-

tainer which itself is bounded-by a Boundary. In-

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

810

Figure 8: Ontology of the rhetorical ﬁgure litotes.

stances of the class of litotes are composed-of in-

stances of Sememe which themselves are part-of the

class of Signiﬁed. Signiﬁed is a linguistic term and

represents the set of concepts (in semantic terms, se-

memes) that are evoked by lexical units. Tying to-

gether the container, signiﬁed and litotes entities is

the UnexcludedMiddle class. We argue that this entity

activates a container (through a neurological process

not understood well and not described herein) and the

speciﬁc order or direction of this relationship is de-

batable.

From image schema theory we say that a container

entails the concept (signiﬁed). This is a metaphoric

action, i.e. signiﬁeds are not actually in anything (ex-

cept, debatably, in our heads), but since we grasp em-

bodied experience intuitively we can use our physical

and spatial understanding of the world to set concepts

in metaphoric bounded containers. Due to the syn-

chrony between containers and signiﬁeds, the unex-

cluded middle class creates a signiﬁed and, further, it

is the unexcluded middle that develops the existence

of the ﬁgure of litotes.

Lastly, we contend that the Unexcluded Middle

is itself invoked by the Contrariety that exists be-

tween sememes (Sememes are subject-to Contrari-

ety). There is further work in this area that should in-

volve the other invoking mechanisms around the Un-

excluded Middle such as the context or knowledge of

the thinker.

The ontology is deliberately underspeciﬁed which

allows for more ﬂexibility in the descriptions we pro-

vide. We have yet to apply this model to many exam-

ples of real-world ﬁgures of litotes, however because

we are modelling at a high level and underspecifying

properties we do not believe this is a problem to be

ﬁxed. We do, however, anticipate that future work

would include greater speciﬁcity and the use of this

ontology in computational applications.

We publish our OWL representation (in XML for-

mat) online

5 CONCLUSION

In this paper we argue that litotes is a rhetorical ﬁgure

that develops a novel middle ground located between

an opposition of terms. The middle ground can be de-

scribed in concept-activation terms and can be consid-

ered in the light of Image Schema theory – especially

the schema of CONTAINMENT. This contrasts with

the logical view of the Law of the Excluded Middle

where concepts are deﬁnitively bounded. Still, there

is a more complex conceptual arrangement involved

where litotes is used and we show a model for the ac-

tivation of new concepts (and Containers) to account

for this novel middle ground(s).

To make this point clearer we must show various

strengths of the ﬁgure – that are reﬂected in the rel-

ative strengths of Implicature. Two further examples

are given, below.

(4) “Not a bad day’s work on the whole,”

he muttered, as he quietly took off his mask,

and his pale, fox-like eyes glittered in the red

glow of the ﬁre. “Not a bad day’s work.”

http://repositori.com/sw/onto/litotes.owl

Cognitive Modeling in Computational Rhetoric: Litotes, Containment and the Unexcluded Middle

811

(Baroness Emmuska Orczy, The Scarlet Pim-

pernel, 1905)

(5) Are you also aware, Mrs. Bueller, that Fer-

ris does not have what we consider to be an

exemplary attendance record? (Jeffrey Jones

as Principal Ed Rooney, Ferris Bueller’s Day

Off, 1986)

In (4) the intent is understatement (a good day’s work

is implied as having been done), but the phrase is a

commonly-used one and usually reﬂects an alterna-

tive and modest way to congratulate someone. In (5),

however, the implication here is that Ferris has a poor

attendance record. It is an understatement again, but

considering that we can picture the scene in the Prin-

cipal’s ofﬁce with all of the context, we might believe

that this second example is a stronger understatement

and reﬂects more weight of intention to make an em-

phatic point about attendance, but without expressly

uttering it.

In summary, litotes creates categorical structures

which map to Containers, and therefore to neural

structures – it is these that give rise to the ability for

litotes to work at all. Furthermore, the balancing of

negated terms versus a contrary – and therefore the

construction of strong or weak opposites – deﬁnes the

ﬁgure of litotes. This balancing force comes about

through the context or wider semantics of the words

in use. Therefore litotes gains strength by the use of

terms with signiﬁcant semantic baggage that allows

the balancing forces to appear.

The power and ubiquity of litotes are evident to

anyone giving modern language usage even just a

brief analysis. This, coupled with the complexity

of the meaning, makes it a very interesting and use-

ful research area. For example in the modern world

of all-pervading social media and a post-truth aspect,

litotes is sometimes used to carry meaning that, due to

its complexity, evades direct computational analysis.

Notably we could look for its presence in hate-speech

in a pejorative sense — which we could call hidden

hate-speech — the speaker is not being hateful in a

direct way, e.g. He is not the smartest pea in the pod.

Our research aims include highlighting the cog-

nitive mechanisms which underlie the ﬁgure in order

to attract greater interest and understanding. Through

more awareness and comprehension we hope to en-

courage more research. A further motivation is to

create computational models of the ﬁgure in order

to drive automated discovery and analysis in a cur-

rently sparse ﬁeld. Our ontological models are pub-

lished online and usable by anyone wanting to per-

form logical inference across linguistic data sets. We

are currently developing further computational mod-

els in this area.

6 FUTURE WORK

Much research has been done in the domain of Image

Schema theory and cognitive understanding, however

there is more to be done (“The cognitive processes

underlying concept invention are still largely unex-

plored ground” (Hedblom et al., 2015)). The perspec-

tive from which we approach these issues is slightly

different and we bring a wealth of rhetorical research

behind us. The following are some areas that we be-

lieve will be important for future research in the cross-

domain territory we are continuing to explore:

1. Our view is that the semantic and pragmatic

weight of a word or phrase can impact the ex-

tent to which conceptual middles can be formed.

We would like to follow this research direction

through exploratory computational corpora anal-

yses.

2. Taking into consideration the characteristics of

litotes and its relation to negativity and Image

Schemata, it is possible to build upon the ontolog-

ical modeling of litotes as a rhetorical ﬁgure, as it

was done in the RetFig project (Mladenovi

c and

Mitrovi

c, 2013) and for the schema of BALANCE

in (O’Reilly and Harris, 2017), and to further use

this new model as a part of argument structure as

envisioned in (Mitrovi

c et al., 2017)

3. Many detailed studies of Image Schemata and

especially CONTAINMENT have been done

(Lakoff and N

nez, 2000; Dewell, 2005; Bennett

and Cialone, 2014; Hedblom et al., 2018). We

aim at research that would combine more of these

deeper theories and alternative methods of under-

standing them with the linguistic domain along

which we have already started.

4. Our ontological representation is deliberately un-

derspeciﬁed which means the deﬁnitions are ﬂex-

ible. This can bring problems in terms of in-

ference, but it can also make the ontology more

pliable to mapping to well-developed upper on-

tologies such as SUMO (Niles and Pease, 2001)

which would allow for inferencing on different

levels of complexity. Our goal is to deﬁne in

more detail the properties of the various entities

we have modeled and to apply this to real-world

examples within computer applications.

REFERENCES

Aristotle (1938). On Interpretation. Harvard University

Press.

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

812

Bennett, B. and Cialone, C. (2014). Corpus guided sense

cluster analysis: a methodology for ontology devel-

opment (with examples from the spatial domain). In

FOIS, pages 213–226.

Blutner, R. et al. (2004). Pragmatics and the lexicon. Hand-

book of pragmatics, 488-514.

Cicero, M. T. (1954). Cicero Ad C. Herennium de ratione

dicendi:(Rhetorica ad Herennium), volume 403. Har-

vard University Press.

Cuccio, V. and Gallese, V. (2018). A peircean account of

concepts: grounding abstraction in phylogeny through

a comparative neuroscientiﬁc perspective. Phil. Trans.

R. Soc. B, 373(1752):20170128.

Dewell, R. (2005). Dynamic patterns of containment. From

perception to meaning: Image schemas in cognitive

linguistics, pages 369–393.

Fillmore, C. J. et al. (1982). Frame semantics. Cognitive

linguistics: Basic readings, pages 373–400.

Frege, G. and Beaney, M. (1997). The Frege Reader. Black-

well Oxford.

Givon, T. (1991). Isomorphism in the grammatical code:

Cognitive and biological considerations. Studies in

Language, 15(1):85–114.

Harris, R., Di Marco, C., Mehlenbacher, A. R., Clapper-

ton, R., Choi, I., Li, I., Ruan, S., and O’Reilly, C.

(2017). A cognitive ontology of rhetorical ﬁgures. In

Proceedings of AISB Annual Convention 2017, pages

228–235.

Harris, R. A. and Di Marco, C. (2017). Rhetorical ﬁgures,

arguments, computation. Argument & Computation,

8(3):211–231.

Hedblom, M., Kutz, O., and Neuhaus, F. (2014). On the

cognitive and logical role of image schemas in com-

putational conceptual blending. CEUR-WS.org.

Hedblom, M. M., Gromann, D., and Kutz, O. (2018). I n, o

ut and through: formalising some dynamic aspects of

the image schema c ontainment. In Proceedings of the

33rd Annual ACM Symposium on Applied Computing,

pages 918–925. ACM.

Hedblom, M. M., Kutz, O., and Neuhaus, F. (2015). Choos-

ing the right path: image schema theory as a founda-

tion for concept invention. Journal of Artiﬁcial Gen-

eral Intelligence, 6(1):21–54.

Hermann, K. M., Grefenstette, E., and Blunsom, P. (2013).

”not bad” is not ”bad”: A distributional account of

negation. Proceedings of the ACL Workshop on Con-

tinuous Vector Space Models and their Composition-

ality.

Horn, L. (2017). Lie-toe-tease: double negatives and un-

excluded middles. Philosophical Studies, 174(1):79–

103.

Johnson, M. (1987). The body in the mind: The bodily

basis of imagination, reason, and meaning. The body

in the mind: the bodily basis of imagination, reason

and meaning.

Kimmel, M. (2005). Culture regained: Situated and com-

pound image schemas. From perception to meaning:

Image schemas in cognitive linguistics, pages 285–

311.

Kuhn, W. (2007). An image-schematic account of spatial

categories. In International Conference on Spatial In-

formation Theory, pages 152–168. Springer.

Lakoff, G. and N

nez, R. E. (2000). Where mathematics

comes from: How the embodied mind brings mathe-

matics into being. Basic Books.

Lanham, R. A. (1991). A handlist of rhetorical terms. Uni-

versity of California Press Berkeley.

Miguel, L. H. (1990). Hoffman, maria, e.:” negatio con-

trarii. a study of latin litotes” (book review). Emerita,

58:346.

Mitrovi

c, J., O’Reilly, C., Mladenovi

c, M., and Handschuh,

S. (2017). Ontological representations of rhetorical

ﬁgures for argument mining. Argument & Computa-

tion, 8(3):267–287.

Mladenovi

c, M. and Mitrovi

c, J. (2013). Ontology of

rhetorical ﬁgures for serbian. In Text, Speech, and Di-

alogue, pages 386–393. Springer Berlin Heidelberg.

Niles, I. and Pease, A. (2001). Towards a standard upper

ontology. In FOIS.

O’Reilly, C. and Harris, R. A. (2017). Antimetabole and

image schemata: Ontological and vector space mod-

els. In JOWO.

Schank, R. C. and Abelson, R. P. (1975). Scripts, plans, and

knowledge. In IJCAI, pages 151–157.

Shovel, M. (2015). Litotes, the most common

rhetorical device you’ve never heard of. https:

//www.theguardian.com/media/mind-your-language/

2015/mar/26/litotes-the-most-common-rhetoric\

\al-device-youve-never-heard-of). Accessed:

2018-10-28.

Taboada, M., Trnavac, R., and Goddard, C. (2017). On be-

ing negative. Corpus Pragmatics, 1:57–76.

Yuan, Y. (2017). The argumentative litotes in the analects.

Argument & Computation, 8:253–266.

Cognitive Modeling in Computational Rhetoric: Litotes, Containment and the Unexcluded Middle

813