=Paper=
{{Paper
|id=Vol-2050/isd_paper3
|storemode=property
|title=Antimetabole and Image Schemata: Ontological and Vector Space Models
|pdfUrl=https://ceur-ws.org/Vol-2050/ISD_paper_3.pdf
|volume=Vol-2050
|authors=Cliff O'Reilly,Randy Allen Harris
|dblpUrl=https://dblp.org/rec/conf/jowo/OReillyH17
}}
==Antimetabole and Image Schemata: Ontological and Vector Space Models
==
https://ceur-ws.org/Vol-2050/ISD_paper_3.pdf
Antimetabole and Image Schemata -
ontological and vector space models
Cliff O’REILLY a and Randy Allen HARRIS b
a Anglia Ruskin University, UK
b University of Waterloo, Canada
AbstractIn this paper we describe the linguistic analysis of the rhetorical figure
Antimetabole. We extend our analysis to cognitive processes, centering on Image
Schemata as a structure of understanding. Representing elements in symmetry as
force vectors we define a graphic representation that shows the balancing forces
in multi-dimensions based on the Image Schema BALANCE. We also describe an
ontology that models the form of the figure and further extends the description to
force vectors in balance.
Keywords. image schema, rhetoric, antimetabole, balance
1. Introduction
In this paper we propose a novel analysis of rhetorical figures from the perspective of
Image Schemata, specifically the schema of BALANCE.
The group of discernible patterns in natural language that have been termed as
rhetorical figures for millennia shine out for their power to cajole and move. These pat-
terns are all around us whether we notice them or not and, in fact, there is no such thing
as “degree zero, purely literal language” [1]. This is well understood for some conceptual
figures known as tropes–particularly metaphor, metonymy, and irony, which Cognitive
Linguistics has revealed are ubiquitous in natural languages. But it is also true of formal
figures, known as schemes, which rely not on conceptual relations but on features of the
signal itself, such as repetition, sequential position, and symmetry. When we recognize
the constituent elements of figures we see that they are frequently present too in visual,
logical, mathematical and conceptual areas of our understanding of the world, e.g. in
cinematic themes [2]. This leads us to value the importance of the element as well as the
gestalt figures themselves. For instance, repetition may contribute to the signal salience
of some element in an utterance, but may also contribute to a more general pattern real-
ized by that utterance. The object of an account that recognizes both individual elements
and gestalt structures is to achieve “the right degree of overlap between the organizing
principles of cognition and the organizing principles of figuration (similarity, contrast,
balance, repetition, and the like)” [1]. We believe that Image Schemata, which, to the best
of our knowledge, have not been utilized in an account of rhetorical schemes, can help
achieve the right degree of overlap, particularly on the cognitive side of the equation.
The rhetorical scheme of Antimetabole is an ancient one recognized in texts as early
as the 6th Dynasty in ancient Egypt and remains ubiquitous in modern discourse, from
advertising jingles and tweets to proverbs and pop songs [3]. Though little noticed (not
unlike many forms of metaphor), occasionally its use flares up enough for special atten-
tion, as it did in the 2012 US Federal Election [4]. It may be defined as the repetition
of at least two words in reverse sequence. A well-known example from Dumas is “all
for one and one for all”. Its importance as a device can be seen from the frequent and
broad range of usage to which it is put and the examples in domains other than that of the
written or spoken word. The general pattern to which the written device conforms also
applies to fields such as mathematics and visual imagery.
Historically analysis of Antimetabole has mainly been within the Arts. Recent activ-
ity has centred on more detailed analytical approaches (e.g. [5] and [6]), but both under-
standing fully the power of this scheme and development of computational techniques
to automatically analyse and discover it leave plenty of work ahead. The key problem
is in elucidating a structured approach to understanding. Our investigations of Image
Schemata intend to bring some cognitive structure to our analyses. This paper describes
developments in a method of representing the balancing forces within an instance of An-
timetabole and also a new ontological description that combines the formal elements of
a figure with the Image Schema BALANCE.
2. Background
Image Schema is a phrase born from the Cognitive Linguistic movement of the latter part
of the 20th century and specifically the work of George Lakoff and Mark Johnson in the
1980s. The word schema itself, however, from the Greek σ χ η̂ µα, meaning “shape” or
“form”, as used in the current context has a provenance more philosophical and psycho-
logical than linguistic — an influence on the choice of this particular word given that
the works of Johnson and Lakoff lie in the fields of cognitive science and philosophy.
It is this original meaning that gives rise to the systematic ambiguity we are attempting
to navigate in this paper, between rhetorical scheme (a figure based on formal features
of language) and image schema (a cognitive entity based on relations that are best re-
alized in formal terms, such as containment and relative position). Immanuel Kant de-
scribes his transcendental schema as a “mediating representation” between “category”
and “appearance” [7] implying a connective role between something innate and some-
thing perceived— in-between a priori and a posteriori knowledge. Piaget’s work on child
psychology influenced a generation of thinkers and continued the idea of a schema as a
constituent of the structural basis of cognition: “schemas are the knowledge structures
that underlie all thinking” [8].
Rudolph Arnheim’s work in the 1970s on the psychology of balance including con-
cepts of force, weight and locus influenced Mark Johnson and George Lakoff who took
this analysis to a deeper level. The idea of a schema as a structural phenomenon contin-
ued, it being “a means of structuring particular experiences schematically, so as to give
order and connectedness to our perceptions and conceptions.”[9]
There are many recognised Image Schemata, but one fundamental example is BAL-
ANCE. Johnson elucidates extensively on this structure and we discuss this further in
subsequent sections: “balance metaphorically interpreted also holds together several as-
pects of our understanding of our world”[9].
Non-computational analyses of the BALANCE image schema are various and range
from expositions of justice [10] to literary analysis [11].
Studies of complex patterns of natural language used to inform or cajole are as
ancient as language itself. A rich history and associated record of rhetorical figures has
descended down the ages so that we presently have a large core of recognised patterns. It
is not a closed set, however, and more investigation into the less well explored languages
and further research on the languages already well understood will surely reveal new and
undiscovered features. In this light it is probable that an exhaustive list is not possible
even in principle (see [6] for a taxonomy). We call the set of these various patterns
Rhetorical Figures which includes very familiar structures such as metaphor (as tropes),
but many more obscure examples from Accumulatio to Zeugma.
Antimetabole is a figure that is well recognized, but not necessarily by its name.
As above, at least two words (or word groupings) repeat in reverse order, often around
a central axis point. A few examples are given below which show particularly well the
balanced formation of words:
If you fail to plan, you plan to fail. (1)
A place for everything and everything in its place. (2)
Fair is foul and foul is fair. (3)
These examples are relatively simple in their structure although there are certainly
complex sets of connotations employed. Another famous example is President John F.
Kennedy’s inaugural address in 1961:
Ask not what your country can do for you - ask what you can do for your country
(Kennedy & Sorensen 1961) (4)
This passage not only reflects the balanced set of words and phrases, but also utilizes
the figure of Antithesis. By adding a negation the balance is thrown out of whack, further
adding to the weight of the intent and the rhetorical power of the passage by nature of
a contrast feature. Mitrovic et al describe Antimetabole in some detail and a number of
ontological approaches to rhetorical scheme modeling [12].
Studies of the cognitive basis for rhetorical figures are not extensive. Gibbs [13]
discusses the understanding of BALANCE and Harris et al [1] propose an ontological
analysis involving descriptions of the form/figure pairings which is a ubiquitous, or even
a necessary, structure of natural language.
2.1. Balance Image Schema
We take Johnson’s work on the BALANCE Image Schema as a basis for our analysis.
For Johnson, the idea of balance is predicated on the basis that cognition is essentially
physical and relates strongly to existing in the world -
“We almost never reflect on the nature and meaning of balance, and yet without it
our physical reality would be utterly chaotic”[9]
As an essential cognitive structure, Johnson describes a prototypical BALANCE
schema in terms of force vectors where forces are relative weights directed around a
point of axis: “Balance involves a symmetrical (or proportional) arrangement of forces
around a point or axis”[9].
Subdivisions of the BALANCE schema given by Johnson are shown in Figure 1
Figure 1. Axis Balance [9]
In Figure 1 plate B shows the prototypical set of vectors centering on a one-
dimensional axis representation. Plates A, C and D in the same figure show Johnson’s
divisions of the prototype in terms of axis variability and the resulting conceptual loci
diversity which is a result of the nature of the elements in balance. Plate A represents
EQUILIBRIUM, plate C is POINT BALANCE and plate D is TWIN-PAN BALANCE.
Some concepts lend themselves to a specific sub-division because the concept itself
seems to fit around the physical constraints of dimensional vectorisation. For example
if we consider the balance of justice we might conceptualize this as a binary division of
truth/falsehood and the corollary guilty/innocent. This binary contrast suits the TWIN-
PAN BALANCE because these concepts, like the physical movements of a weighing
scales, vary in weight in two possible directions (up or down) and with two possible
subjects being compared.
Johnson further analyses the BALANCE schema away from the vector model and
creates a set of types:
1. Systemic Balance – the perceived proper balance of forces implied by a func-
tional unity, e.g. a system such as weather which we might perceive as balanced
by rainfall and sunshine
2. Psychological Balance – the structure of our experience concerning the “balance
of intellectual, physical, social, religious and model activity”
3. Balance of Rational Argument – comparing opposing sides of an argument as
balance by virtue of the weight of evidence
4. Legal/Moral Balance – a concept of a natural order being upset and imbalanced
is re-weighed and put back into proper balance according to argument, force and
weight
5. Mathematical Equality – a mapping from the physical into the abstract mathe-
matical so that equations exist and can be balanced in terms of abstract concepts
(numbers, algebraic terms etc)
These types define conceptual dimensions within which things in the world are able
to be balanced as perceived by observers. Each one relates to a set of concepts and asso-
ciated meanings in common usage. Roughly, we can refer to these notions as noetic bal-
ance. These are higher order conceptual arrangements that reflect states of knowledge (an
argument may be balanced or unbalanced, a climate and an ecology may be in balance or
out of balance, and so on). While these types of balance are often expressed through lin-
guistic patterns, as we will see for some of them, they are not our direct concern. It is the
formal linguistic patterns themselves that matter for an account of rhetorical schemes.
Ultimately what is described as in-balance encompasses elements inferred and ab-
stracted by the perceiver. From metaphorical projections we can say that related ideas
such as “psychological states, arguments, moral rights and mathematical operations” are
in balance as much as the orthographic words on the page that invoke them.
2.2. Antimetabole
One for all and all for one
— Dumas, The Three Musketeers (5)
For y am sorwe, and sorwe ys y
— Chaucer, The Book of the Duchess (6)
This “poorly researched and largely unknown but remarkably widespread rhetorical
device” [1] is characterised by a repetition of elements around a central axis, but with
a reversal of the positions of the elements which gives rise to a symmetry. Often sym-
bolised as ABBA (A and B referring to elements in the text - usually words or word
groupings). There are many examples of this pattern in natural language, but the form
also pervades in mathematics, logic and visual phenomena. A symmetry of multiple parts
is not uncommon in any aspect of life, however Antimetabole is different in the sense
that it is linguistic and therefore encourages semantic and pragmatic responses that cre-
ate a gestalt structure which is both memorable (for its repetition) and powerful (for its
constructive inference). Antimatebole belongs to the super-set of figures called chiasmus
(from the Greek letter chi (χ) for the literal criss-cross nature of the character’s form)
which includes any figure that has a reverse repetition structure.
A key method for analysing the meanings of natural language and especially useful
in understanding rhetorical forms is to look for form/function pairs. Decoupling the way
a word or phrase appears (form) from the way it evokes meaning (function) allows us
to focus separately on these items and remove any confusion in a perceived overlap. For
example, we might say that the form of the word “all” from Dumas’ example above is a
noun representation with two occurrences varying between object and subject grammat-
ical roles, while “one” a noun representation with two occurrences varying between sub-
ject and object grammatical roles, and that their positions relative to each other (as well
as their respective grammatical roles), pivot on a single axis word (“for”); meanwhile
the two proximal phrases (“one for all,” “all for one”) pivot on another single axis word
(“and”). The function of the “all” could include the ideas of collaboration and support
from an individual’s contribution to the group effort and also from the group’s support of
individuals.
Antimetabole always features balance. The repetition of elements around an axis
always gives two or more features that act in symmetry somehow. Antimetabole is com-
plex, however, and the elements in balance can range from simple words to more complex
concepts. We describe these different levels of balancing elements in terms of Johnson’s
division into types as described in the previous section.
Ruan et al [3] describe the cognitive affinities evoked by this pattern as OPPOSI-
TION, REPETITION and SYMMETRY. Further detailed analysis shows the underlying
structure as containing these essential functions:
1. QUANTITY: the number of repeating elements
2. SEQUENCE: an ordering of repeating elements is essential
(a) The ORDER of the sequence gives rise to semantic functions of IRRELE-
VANCE OF PRIORITY and COMPLETENESS
(b) The DIRECTIONALITY of the sequence gives rise to RECIPROCALITY
and OPPOSITION
3. PROXIMITY: elements must be proximal which gives rise to a “joint conceptual
force”
Harris further argues that a more thorough investigation of rhetorical forms such
as Antimetabole is a continuation of the “renewed interest in tropes, such as metaphor”
which, famously, located a rich seam of understanding through the field of Cognitive
Linguistics. We believe that the melding of rhetorical analyses with established theories
in Cognitive Science (i.e. Image Schemata) is of potentially fruitful benefit to the devel-
opment in our understanding of animal cognition. The process of viewing that which is
ubiquitous,but not well understood through a lens of cognitive analysis has worked well
in the past (e.g. metaphor) and we have no reason to think it would not continue to bring
benefits with the combination of other rhetorical schemes and image schemata.
3. Methods
Our research focuses on the multi-dimensional balances brought about conceptually by
the figure of Antimetabole. We take inspiration from the Image Schema theory and seg-
ment the linguistic information – semantic and pragmatic – gleaned from a particular
form and put these into a multi-dimensional model. We devise a graphing output that
highlights the balances involved in a figure and also propose a numerical vector-based
model for future work.
Using the same premise of multi-dimensional balances, we devise an ontological
approach that specifies the elements in a computationally-accessible format via the On-
tology Web Language (OWL).
3.1. Analysis
Our intention is to analyse various texts for inherent elements in balance according to the
types (or as we define them, dimensions) from Johnson, as described above. We proceed
by example in the analysis of two figures — from Dumas and Kennedy. Dumas’ famous
saying “all for one and one for all” obviously contains a surface repetition of two words
- “all” and “one”. It also has a pivot point between the first “one” and second “one”. The
balance evident between the written words we consider as a Systemic Balance — when
we see words that are the same we notice the similarity and perceive a force relationship
between each word and the axis. There are multiple levels of balancing forces in play.
An individual word may be considered balanced with another, but also there is a balance
of word groupings, i.e. not only do the two instances of “all” balance, but the phrases
“all for one” and “one for all” can be said to pivot around the central word of “and”. We
know too that the meaning of the words in this saying is related to duty and responsibility
— if we know the context of the story from which it is taken. As we bring into focus the
further interpretations of the words — concepts of support and contribution — we move
from a Systemic Balance to a Psychological Balance. Finally we might see that there is a
larger interpretation of the duty to one’s companions and a responsibility to one’s cause.
This is the Moral/Legal Balance as an extension to the Systemic and Psychological.
Looking at Kennedy’s famous quote from his 1961 address: “Ask not what your
country can do for you - ask what you can do for your country” we see, on the surface,
orthographic words that repeat – “your country”, “can do for”, “you”. We see similar
structures as in the Dumas example and, again, the repetition leads us to pinpoint a central
axis about which the repetition occurs. This isn’t always a precise operation, but in this
example we take the division between words 9 and 10 (“you” and “ask”) as the axis. We
take the orthographic repetition to be a Systemic Balance.
When we delve deeper into the figure we can see the underlying meaning of the
passage (some of which is sketched out by a number of political commentaries e.g. [14]).
The writer’s intention is not simply a repetition of words for effect, but a contrasting and
balancing set of forces that result in concepts such as duty, Government, rights etc.
We infer from the phrases “your country” and “you” a metaphoric transfer to
concepts of Government and Individual respectively. We understand the context of
Kennedy’s address and the Systemic Balance inherent in concepts such as country, lead-
ership and individual etc. We show this as an imbalance in favour of Government —
from a simplistic assessment of the physical and metaphorical extent of the two.
We consider further the implications of the passage. In effect the idea of the indi-
vidual taking from the state (emphasised by the negation in the first phrase) is contrasted
with the idea of the individual giving to the state. We add “taking from” and “giving to”
as balanced concepts. We move away from a Systemic Balance and into Psychological
Balance at this level.
Finally we extrapolate to the concepts of “rights” and “responsibilities”. These are
elements in balance around the Legal/Moral Balance type.
Our interpretation and mapping to schema type is not a definitive assessment - many
other interpretations are possible.
We show in Figure 2 a graph of some of the concepts and forces evoked in this sec-
ond rhetorical figure. Directed lines are vectors with length representative of the rela-
tive weight of concept. In this format we can see the balances and imbalances resultant
from the analysis. We also show relationships between elements as dotted lines. These
don’t play a direct role in the relatively simple balance relationships we document, but
are important for the conceptual analysis. A quantity value must be assigned to the vec-
tors in order to use them mathematically. The nature of calculations that would provide
these numbers is beyond the scope of this paper, but within the scope of future work.
The elements in balance and the calculations that derive vectors from them are open
to interpretation and very much subjective, however we believe that consensus can be
gained on many areas and, where this is not possible, a publishing system that allows the
co-existence of different perspectives would be sensible – this is, after all, how humans
often see the same thing.
Figure 2. Kennedy’s Antimetabole - BALANCE interpretation graph
Further utilising Johnson’s description of factors – implied by an Image Schematic
analysis as vectors – we describe a multi-dimensional vector space as a way to imagine
and calculate a particular Antimetabole figure. Our inspiration is Johnson’s analysis and
our motivation is the potential for intuitive visualisation and the power of mathematical
analysis. We use a simple Euclidean Vector with magnitude and direction to indicate the
position and weight of the term relative to our point of axis. Our dimensions relate to the
types described above, but are certainly not limited to these. The output so far is restricted
by practicality: our imagination and the ability to represent greater than 3 dimensions in
a diagram. Figure 3 shows two graphs that represent a hypothetical analysis of balanced
elements from Kennedy’s speech. Each vector is directed at the axis (0,0 in the graph), but
has differing direction depending on the dimensional weights associated to the particular
element that asserts the vector. The vectors that arise from a figure would be combined
into a multi-dimensional vector space.
We don’t necessarily include mathematical or logical assumptions of vector space
analysis in our model, but these could be avenues for inclusion in the future in order
to take advantage of useful techniques such as Vector Algebra or Support Vector Ma-
chines. Mathmetising a collection of vectors that represent the various elements within
a passage of natural language is not new, but to do so with in the cognitive framework
of Image Schema theory would be worth investigating. The benefits of vector analytics
to language analysis is significant ([15] [16] [17]) and one potential research goal would
be in rhetorical scheme categorisation. The maths is relatively straightforward, but the
determination of the content of vectors is as complex as computational rhetoric can be
and we leave that to future research.
Our focus in this paper is Antimetabole and BALANCE, however we foresee
other figures being profitably viewed through the perspective of BALANCE and other
schemata. Figures such as Anadiplosis, Paralellism and Epanalepsis exhibit elements
of balance, and other schemata such as SOURCE-PATH-GOAL can relate to Climax or
Amplification, while Digressio has elements related to DIVERSION.
Figure 3. 2-dimensional vector space representations of balance forces
3.2. Ontology
We describe a novel ontology of the balance components of the rhetorical figure An-
timetabole with reference to the form and function of constituent elements and also con-
taining a model of the cognitive affinities involved and their relation to the Image Schema
of BALANCE1 . The purpose of developing this ontology is manifold:
• to provide a structural reference for future analyses including computational ap-
proaches
• to generate greater understanding of the figure by the process of ontological anal-
ysis
• to gain insights into the cognitive processes underpinning the conceptualisation
of this figure
Harris et al 2017 [1] describe the Rhetfig project that has analysed this particular
figure for the composition of an OWL ontology. The authors describe the various ap-
proaches to modelling a complex linguistic structure such as Antimetabole. Their anal-
ysis calls the conceptual elements involved in rhetorical figures Cognitive Affinities
such as CONTRAST, SIMILARITY, SEQUENCE, REPETITION and POSITION. An-
timetabole, by their analysis, can utilize the affinities of REPETITION, SEQUENCE and
CONTRAST. In order to describe these affinities, our ontology must contain classes for
the basic elements that are repeated, in sequence and that contrast each other.
As stated previously, when we define the concept of balance as a repetition around a
point then we can see that balance is a feature of the Antimetabole figure. Especially rel-
evant is the SEQUENCE affinity which describes the ordering of the elements in the fig-
ure and therefore the location of the focal point of the balance structure. In the symbolic
1 Available at http://repositori.com/sw/onto/antimetabolebalance.owl
representation A-B-B-A, the axis must exist in-between the B symbols for the symmetry
to work and the distances from the axis of each element are comparable and equal — the
As are equidistant (and so are the Bs) from the perceived axis.
Figure 4 shows a graph of the ontology for the rhetorical figure of Antimetabole
and the BALANCE Image Schema. Various classes are present that describe the figural
form, e.g. Word and Similarity. The Word class is central to the ontology and affects
the concept of a Force Vector which in turn evokes the Balance class. The key concepts
we attempt to specify are Axis, Vector and Word: these three concepts form the centre
of a structure that exemplifies the things that are in balance and about which they are
in harmony and/or tension. Further, it conceptualises a sense of a directed force which
evokes the vector model and also relates to Johnson’s foundational work. Our graph is
underspecified at present and we may attempt to define the details further if required by
the models that take it as a basis for analysis.
Figure 4. Ontology graph - combining Rhetorical Figures with Image Schema
We add logic rules around this ontology in order to define knowledge of a particular
instance of Antimetabole and its sub-structure. For example, the SWRL2 rule below (7)
produces the instance of Balance between two Words when they evoke Force Vectors of
the same Dimension and directed at the same Axis.
Axis(?a1) ∧Word(?w1) ∧ ForceVector(? f v1) ∧ evokes(?w1, ? f v1)
∧Word(?w2) ∧ ForceVector(? f v2) ∧ evokes(?w2, ? f v2)
∧directed − at(? f v1, ?a1) ∧ directed − at(? f v2, ?a1)
∧Dimension(?d1) ∧ o f (? f v1, ?d1) ∧ o f (? f v2, ?d1)
→ Balance(?b1) ∧ evokes(? f v1, ?b1) ∧ connects(?b1, ? f v2) (7)
2 https://www.w3.org/Submission/SWRL/
An important goal with ontology engineering in the domain of language analysis is
to be able to share and re-use the work of others. By publishing ontologies and knowledge
bases to the internet we hope to encourage others both inside and outside of academia
to benefit from agreed definitions for shared concepts. This ontology is represented in
OWL (Ontology Web Language) and therefore useful for integration into the Semantic
Web or other computational applications that can utilise XML representations.
Evaluating our approach takes a number of forms; we consider both the output and
the process. The process of analysis has brought new understanding and shone a light
on new pathways of discovery yet to be followed. The ontology as an output is not
yet evaluated, but future work will include using the OWL descriptions in action for
computational processing.
4. Discussion
Our goal is to elucidate structure to the understanding of Antimetabole. We attempt this
through a novel method for analysing rhetorical figures, a new graphical representation,
a route to future mathematisation, and an ontological framework that is computationally
tractable and re-usable. The core of our analyses is the principle that things in the world
are readily balanced and that human sensibilities (as Image Schemata) are fundamental
to cognitive processes and can be modelled into complex patterns of natural language
(rhetorical figures).
We have attempted to sketch the possibility of a strong connection between Image
Schemata and Rhetorical figures. There is rich pickings in these areas and we scope some
future directions for research, below.
Aside from analyses of metaphor in relation to Image Schemata, we believe this
research is novel in its elucidation of rhetorical figures from the perspective of Image
Schema theory. Others have investigated cognitive processes in poetry and allegory, for
example (e.g. [11][18]) and many others the computational and cognitive dimensions of
rhetoric (e.g. [5][1]), but we have pushed this general theme into a new direction. We
recognise that the picture we paint is limited to one particular figure and one particular
schema, but we expect that there are many more areas to be investigated, as described
briefly above, that will include the diverse and extensive range of rhetorical figures and
include more schemata beyond that of BALANCE.
We hold to the idea that ontology engineering often brings benefits both in terms of
the eventual output (e.g. as an XML representation to be shared and utilised), but also
for the process itself of analysing a particular domain. This has been a theme of the work
in analysing rhetorical schemes where significant insights have arisen from ontological
activities. This is not the only goal, however, and we aspire to take all ontologies forward
into computer models that do a number of different tasks from describing, quantifying,
discovering and elucidating on what is a fascinating and important domain of artificial
intelligence research.
5. Future Work
We hope that a number of areas for further investigation are evident after the work de-
scribed in this paper. Arguably the ultimate goal of a cognitive analysis of language is to
better understand its workings and to shine a light on a further, deeper understanding and
perhaps even automation through Artificial Intelligence; we are still a significant distance
from being able to fully comprehend what creatures are doing with their brains when
they communicate. A perhaps unforeseen outcome of deep investigations of language is
the difficulty in agreeing on what things mean. There are multitudinous interpretations
of any span of text (which is to be expected and encouraged), but singular semantics can
be valuable and through consensus between humans in concert with computer models
we can hope to add important knowledge to shared resources.
We highlight some areas we believe would be fruitful for further analysis, below:
1. In the work described in this paper we focus wholly on the figure of An-
timetabole. There are many others that rely on the concept of balance and we
would like to continue this analysis to figures such as Anadiplosis (“The moun-
tains look on Marathon And Marathon looks on the sea” [Byron] or Antithesis
(“That’s one small step for man, one giant leap for mankind. [Neil Armstrong]”)
2. We have also focussed solely on the Image Schema of BALANCE. There are
a host of important schemata that we would very much like to investigate from
Containment to Source-Path-Goal etc. We believe that much more interesting and
useful understanding will be gathered by looking in more detail across the range
of conceptual functions
3. We only barely touch on the power of the metaphorical transfer to Force Vectors
and mathematising the concept space of a rhetorical figure. Vector mathematics
is a mature set of structures for representing abstract and concrete entities and
we expect that future work in this direction would be extremely interesting. One
interesting avenue we would like to see is a mathematical comparison of figures
in vector terms. We appreciate that the formation of the precise vectors within a
specific figure is open to interpretation, but given the right framework we imagine
it would be possible and very useful to compare individual examples of figures
in quantitative terms.
4. When we described the example from Kennedy we modelled the relationships
between elements simply as dotted lines. We didn’t take them as part of the bal-
ancing structure. This was a necessary move for clarity, but we believe that these
relationships are important and it is likely that by analysing them we will be able
to extend the ontology and vector models accordingly.
5. We have talked about the benefits of the ontological approach and the understand-
ing that comes from simply thinking hard about how certain kinds of language
(and cognition) work. We value these benefits, but also realise that evaluation is
important. A next step must be to structure the over-arching goals of Computa-
tional Rhetoric so that a consistent and measurable framework can be drawn-up
from which even greater advances can be created.
References
[1] Harris RA, Di Marco C, Mehlenbacher AR, Clapperton R, Choi I, Li I, et al. A Cognitive Ontology of
Rhetorical Figures. In: Proceedings of AISB Annual Convention 2017; 2017. p. 228–235.
[2] Coëgnarts M, Kravanja P. Embodied visual meaning: Image schemas in film. Projections. 2012;6(2):84–
101.
[3] Ruan S, Di Marco C, Harris RA. Rhetorical Figure Annotation with XML. In: Computational Models of
Natural Argumentation (CMNA)16, A Workshop at the 2016 International Joint Conference on Artificial
Intelligence (IJCAI). New York; 2016. .
[4] Lapidos J. The Hottest Rhetorical Device of Campaign ’08;. Accessed: 2017-07-06.
http://www.slate.com/articles/life/
the good word/2008/09/the hottest rhetorical device of campaign 08.html.
[5] Dubremetz M, Nivre J. Machine Learning for Rhetorical Figure Detection: More Chiasmus with Less
Annotation. In: Proceedings of the 21st Nordic Conference on Computational Linguistics, NoDaLiDa,
22-24 May 2017, Gothenburg, Sweden. 131. Linkping University Electronic Press, Linkpings univer-
sitet; 2017. p. 37–45.
[6] Harris R, Di Marco C. Introduction to the special issue on Rhetorical Figures. Argument and Computa-
tion. 2018;7(3).
[7] Kant I. Critique of Pure Reason. Translated by Norman Kemp Smith. London Macmillan 1934.; 1934.
[8] Piaget J, Cook M. The origins of intelligence in children. vol. 8. International Universities Press New
York; 1952.
[9] Johnson M. 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. 1987;.
[10] Antle AN, Corness G, Bevans A. Springboard: Designing image schema based embodied interaction
for an abstract domain. In: Whole Body Interaction. Springer; 2011. p. 7–18.
[11] Freeman MH. Cognitive linguistic approaches to literary studies: State of the art in cognitive poetics.
In: Geeraerts D, Cuyckens H, editors. THE OXFORD HANDBOOK OF COGNITIVE LINGUISTICS.
Oxford University Press; 2007. p. 1175–1202.
[12] Mitrovic J, O’Reilly C, Mladenovic M, Handschuh S. Ontological Representations of Rhetorical Figures
for Argument Mining. Argument and Computation. 2018;7(3).
[13] Gibbs RW. The psychological status of image schemas. From perception to meaning: Image schemas
in cognitive linguistics. 2005;29:113–136.
[14] Vanatter S. [Website resource] Dissecting Kennedys Ask not from his first inaugural address; 2013.
Accessed: 2017-07-11. https://www.ff.org/dissecting-kennedys-ask-not/.
[15] Reitter D. Simple signals for complex rhetorics: On rhetorical analysis with rich-feature support vector
models. In: LDV Forum. vol. 18; 2003. p. 38–52.
[16] Pang B, Lee L, Vaithyanathan S. Thumbs up?: sentiment classification using machine learning tech-
niques. In: Proceedings of the ACL-02 conference on Empirical methods in natural language processing-
Volume 10. Association for Computational Linguistics; 2002. p. 79–86.
[17] Sidorov G, Velasquez F, Stamatatos E, Gelbukh A, Chanona-Hernández L. Syntactic n-grams
as machine learning features for natural language processing. Expert Systems with Applications.
2014;41(3):853–860.
[18] Sinding M. Assembling spaces: The conceptual structure of allegory. Style. 2002;36(3):503–522.