Ploke is a rhetorical device of lexical repetition, with multiple variations contingent on place of occurrence. It is widespread in all natural and artificial languages because it manages stability of reference and predication. Syllogisms, for instance, are heavily dependent on positional repetition. Ploke also influences the reader's/hearer's attention because of its appeal to neurocognitive afinities. A formal knowledge representation of ploke is therefore valuable for any AI/NLP system. This paper proposes an ontological model for ploke. We discuss components of diferent types of plokes and rhetorical figures in general, in terms of their form, their function, and the associated neurocognitive afinities that afect attention.
Rhetoric is the “ancient study of persuasion, with particular attention to the efects of expressive
style on belief, action, and knowledge” [
Devices of rhetoric called rhetorical figures which generate a set of attentional efects such as
salience, aesthetic pleasure, and mnemonic efect received by the audience, thus managing the
receiver’s attention. Common rhetorical figures, such as rhyme, metaphor, and sarcasm, that
are encountered in everyday conversations may be noticed easily. Other common rhetorical
ifgures that involve traits, such as repetition, semantic increase, and inverted concepts, may not
be as easily noticed. A rhetorical figure creates a “figure-and-ground landscape that focalizes
some details and put others in relief” [
However, few of even the most ubiquitous rhetorical figures have received attention from computer scientists (or even rhetoricians). One such example is ploke1, rhetorical figures of lexical repetition. That is, whenever a word, word string, or entire phrases or clauses is repeated, it is a ploke, e.g., “very, very big”; “easy come, easy go”; “waste not, want not”. Rhetorical figures code much valuable information at syntactic, semantic, and pragmatic levels. This information can be incorporated to facilitate a wide range of AI/NLP applications such as conversational agents, machine translation, authorship attribution, genre detection, sentiment analysis, etc. This paper proposes an ontology for ploke and other rhetorical figures of lexical repetition in terms of class, subclass, and their relations. We also represent concepts related to ploke and rhetorical figures in general, e.g., form, function, neurocognitive afinity, etc., Section 2 describes the current state of research and applications of rhetorical figures in AI/NLP. Then, Section 3 presents classes and relations that capture the high level concepts of rhetorical figures in general. Section 4 discusses detailed concepts related to ploke in ontological terms. Section 5 concludes the paper with a discussion potential future research directions.
Although rhetorical figures have been studied for millennia, the establishment of the subfield
Computational Rhetoric was fairly recent, where the importance of a computational model
for rhetorical figures was highlighted for various NLP applications such as argument mining,
authorship attribution, automated rhetorical figure detection, etc [
Ontologies are formal knowledge representation models that enable sharing of reusable and
inferable knowledge structures and data. The most cited definition of an ontology is from
Gruber: “An ontology is a formal, explicit specification of a shared conceptualization” [
In previous work of O’Reily et al. [
The goal of this paper is not about the details of a detection algorithm for rhetorical figures, but
rather to develop a unified ontological representation that facilitates various NLP applications,
one of which is the automatic detection and annotation for rhetorical figures. Existing detection
tools uses methods such as regular expressions [18], shallow parsing [
This section describes a general level ontological model for rhetorical figures. The ontology includes a classification of rhetorical figures as well as related concepts such as linguistic elements, neurocognitive afinities and attentional efects. This model is not intended to be comprehensive but rather as a collection of upper level classes and relations that encapsulate concepts of ploke presented in the next section. We use a capitalized word in a sans serif font to indicate a class name, e.g., Ploke; a capitalized word in a typewriter font to indicate a relation, e.g., Contain; a serif font with quotation marks for an example quotation of a rhetorical figure or a part thereof, e.g., “All for one”; and text in a serif italic font for a variable denoting an arbitrary instance of a class, e.g., passage1.
To the center of the ontology is the class Rhetorical Figure. A Rhetorical Figure, contained in a Passage, causes the Passage to have Attention Efect s on the Passage’s Receiver, as is shown by the bolded path in Figure 1. A conventional classification of rhetorical figures is to classify ifgures as schemes and tropes. A Scheme is a subclass of Rhetorical Figure that is characterized by form such as phonology, morphology, lexis, and syntax. A scheme can be formed by, for instance, an efective repetition or alternation of position of a word or phrase in a passage. A Trope is a class in which the figures are heavily characterized by the semantics within the passage. The following example demonstrates an instance of Scheme: “All for one, one for all”[23] (Example 1)
Example 1 contains multiple rhetorical figures, including several plokes. A Ploke is a subclass of Scheme whose form is a repetition of words. In this example, the repeated words are “all”, “for”, and “one”. While Example 1 is most famous as an exemplum of antimetabole, reverse lexical repetition (“all” and “one” repeat in reverse order), rhetorical figures often work together to maximize cognitive appeal and constrain the range of rhetorical functions [24], and Example 0,*
Linguistic Element Token 1 is no exception. It consists a number of diferent types of scheme in addition to antimetabole, including mesodiplosis, anadiplosis, and others, which will be discussed in the next section.
A Rhetorical Figure consists of a Form and a Function2. The Form of a Rhetorical Figure
consists of a set, possibly ordered, of Linguistic Elements. A linguistic element is one of a word,
syllable, phrase, clause, phoneme, lexeme, morpheme, sememe. Since this paper focuses on
ploke, we will omit details of other linguistic elements and focus our discussion on Token only,
which is defined in this paper as one word or a sequence of words, minimally interrupted. Token
is a subclass of Linguistic Elements. The Form of a Rhetorical Figure Triggers one or more
neurocognitive afinities. A Neurocognitive Afinity is a pattern recognized by a receiver’s brain
[
The connection made by the neurocognitive afinities creates a coupling between a rhetorical
ifgure’s form and its function. The function becomes predicable when the form of a rhetorical
ifgure is recognized [
The terminological and conceptual history of rhetorical figures is uneven at best, and ploke
is a case in point. It is traditionally defined as “the intermittent or unpatterned reappearance
of a word” [27], distinguished from figures like antimetabole or anadiplosis, in which words
do exhibit patterns in their occurrences. But as Harris [
“. . . you won’t prevent poverty. Poverty being more than income . . . ” [21]
“Ask not what your country can do for you. Ask what you can do for your country.” [29]
“Boys will be boys.”
“That’s one small step for man, one giant leap for mankind.” [30]
“If I lend you money, I’m taking a risk. And I should carry that risk.” [21]
“Long, long ago.”
“Man behaving strangely is confronted by police. Police shoot man. Man dies.” [31]
“We were elected to change Washington, and we let Washington change us.” [32] The use of these rhetorical figures generates a set of attentional efects linked to a number of neurocognitive afinities of human mind such as balance and symmetry, and to its core, repetition and position.
Recall that a Rhetorical Figure consists of a Form and a Function. Ploke as a subclass of Scheme which in turn is a subclass of Rhetorical Figure, inherits this property. The form of a ploke, i.e., PlokeForm, a subclass of Form, Triggers a set of subclasses of Neurocognitive Afinity such as Repetition, Position, Balance, Opposition, autc. Each subclass of Ploke has a specicfi form which triggers a set of neurocognitive afinities. For example, the form of an antimetabole such as Example 1 is: where and are Tokens being repeated ( = “all” and = “one”). The form of antimetabole demonstrates a balanced structure with the reversed repetition of and , therefore triggering neurocognitive afinities of a reversed repetition which in turn contributes to afinities such as balance and opposition. We will discuss the relations between forms and neurocognitive afinities with details in the next section.
The most important functions of ploke are to stabilize reference and emphasize the concept conveyed by the repetition. Repetition of a Token in a passage or speech stabilizes the reference to increase its clarity in context. Fahnestock’s analysis suggests that repetitions are crucial in scientific writings where it is important to reduce or eliminate ambiguity [ 25]. Another function of ploke is to emphasize the concept conveyed by the repeated Token. The more a word, phrase or clause is repeated, the more it stands out from a sentence or a passage. Both stabilizing reference and emphasizing concept conveyed are functions related to ploke. In addition, subclasses of ploke have specific functions such as to increase the significance of quantity, reciprocality, and irrelevance of order, etc. which will be discussed in the following paragraphs.
An antimetabole has several functions. A function of antimetabole is to convey the idea of irrelevance of order, that is, the order or rank of and does not matter. For example, “Ladies Linguistic Element Token hasLE hasForm
Ploke and gentlemen, gentlemen and ladies”. Another function of an antimetabole is to generate a focus on reciprocal relationship [25]. The reversed repetition often suggests a reciprocal relation between the repeated words and . For example, in propositional logic, the equivalence relation ↔ is expressed as: → ∧ → , which indeed is an antimetabole, i.e., it has the form of . . . . . . . . . . . . . . .. More generally, let 1 and 2 be binary relations, and be atomic concepts, then a reciprocal relation between and can be represented as where the * can be any binary connective, e.g., ∧, ∨, →, or ↔, which does not afect the antimetabole. This is represented graphically in Figure 3.
1(, ) * 2(, ) ← 1 2 →
In Example 1, the reversed repetition suggests an equivalent reciprocal obligation between “all” and “one”, i.e., an entity of group, organization, or community supports an individual while an individual contributes to that entity. Thus, let For be a binary relation, we have
For(“all”, “one”) → For(“one”, “all”)
Note that the symmetrical case in which 1 = 2 is achieved in combination with mesodiplosis, i.e., the medial repetition of the word “for”. The function of irrelevance of order or rank also applies in Example 1, thus the rank of “one” and “all” does not matter. Therefore Example 1 may also be expressed as:
“One for all, all for one” Combining the expression above with the original Example 1 results in an equivalence relation
For(“all”, “one”) ↔ For(“one”, “all”) Consider the following example in which exists between and but not equivalent. “Anyone who thinks he has a solution does not comprehend the problem and anyone who comprehends the problem does not have a solution.” [33] (Example 2) where = “has a solution” and = “comprehends the problem”. Converting it to propositional logic we have,
( → ¬) ∧ ( → ¬)
We cannot conclude the opposite, i.e., ¬ → , anyone who does not comprehend the problem has a solution. The formula above also resembles the form of an antimetabole. If we abuse the notation and define an arbitrary relation IF-Then-Not such that
IF-Then-Not(, ) ≡ → ¬ Thus, the reciprocal relation conveyed is the IF-Then-Not relation between and .
IF-Then-Not(, ) ∧ IF-Then-Not(, )
Note that IF-Then-Not(, ) ≡ IF-Then-Not(, ) since ( → ¬) ≡ ( → ¬), and can be reduced to ¬( ∧ ). Example 2 basically repeated ¬( ∧ ) in diferent forms, one starts from and ends at , one starts from and ends at . The function of completeness is evoked by the antimetabole in the sense that: doesn’t matter where you start, whether you think you have a solution or you comprehend the problem, a solution does not exist.
IF-Then-Not(, ) ≡ ¬ ( ∧ ) ≡ IF-Then-Not(, )
An antimetabole usually conveys a relation 1 and 2 between and . Usually a mesodiplosis suggests that 1 = 2. Recall the formula with an abstract binary connective, The binary relation * can be simple logical connectives such as ∨, ∧, →, ↔. In some cases * is explicit as in Newton’s third law of motion, i.e., “If you press a stone with your finger, the finger is also pressed by the stone” [34]; and in other cases it is implicit as in Example 1 and Example 2. The functions of plokes, or any rhetorical scheme, are best achieved when accompanied by forms that evoke or trigger neurocognitive afinities that harmonize with the functions.
Repetition is an important and fundamental aspect of language and cognition. It is possibly
the most fundamental neurocognitive afinity that builds the foundation of brain activities and
thus human minds [
Another example is the form of anadiplosis, i.e., . . . . . . . , which generates a focus point
that serves as a transition point of a step-wise energy flow as shown in Figure 4b. The repetition
at the final position of the boundary acts as the goal of the first clause, then the repetition at
the initial position of the successive clause acts as the source of that clause [
The position afinity is triggered by repetitions of a Token at diferent positions within a phrase or clause. Positions can be immediately followed as in an epizeuxis, reversed as in an antimetaoble, or proximal with respect to another token in a passage. A position can also be the initial, medial, or final position with respect to a clause with in a passage.
Traditionally, neurocognitive afinities are viewed as independent entities. However, following the ontological representation of concepts related to Ploke presented in this paper, it is reasonable to view repetition and position as two primary neurocognitive afinities triggered by the form of a ploke. That is, Repetition and Position are subclasses of Neurocognitive Afinity . All other neurocognitive afinities triggered by forms of subclasses of Ploke are subclasses of Repetition linked with specific orientation of positions that are subclasses of Position.
Repetition is the reoccurrence of an Element. The Element in this context may be referred to as a Token, phrase, clause, or a repetition, e.g., figures such as antimetabole and gradatio have nested repetitions. A Repetition consists of at least two Elements that refer to the same instance of a Token. In this case, an Element can be viewed as an occurrence of a Token. An Element has a Position which can be Before or After another Position. Thus the Elements’ Positions impose some order, i.e., Before(position(e1), position(e2)) for Elements e1 and e2. Note that position(e1) and position(e2) are functions which are in turn instances of the Position class. A Repetition can be defined as follows: ∀rep, e1, e2.Repetition(rep) ∧ Element(e1) ∧ Element(e2) ∧ ElementOf(e1, rep) ∧ ElementOf(e2, rep) → RefersTo(e1, w) ∧ RefersTo(e2, w) ∧ Token(w) ∧ Before(position(e1), position(e2))
A Repetition can be an Unpatterned Repetition, Relative Repetition, Respective Repetition, or a Nested Repetition. An unpatterned repetition is triggered by the form of a Conduplicatio, a sublcass of Simple Ploke, in which repetitions are not constrained by position. Other subclasses of Complex Ploke such as Epanaphora, Mesodiplosis, and Anadiplosis further restrict positions of their elements with respect to the clauses that contain them4. The form of each of these ifgures triggers Constituent-Initial, Constituent-Medial, and Constituent-Final Repetition respectively, all of which are subclasses of Relative Repetition. The form of Epizeuxis tirggers Immediate Repetition, whille the form of an Antimetabole triggers Reversed Repeition. Both are subclasses of Respective Repetition. Organization of sublcasses of Repetition is shown in Figure 5.
Token Unpatterned Repetition
Position
In Example 1, repetition of the word “all” can be represented by referring both e1 and e2 to the word “all” as in Figure 6. Similarly, the repetition of the word “one” can be represented using diferent sets of instances of Element referring to the word “one”. Both repetitions of “all” and “one” are triggered by forms of subclasses of Ploke, more specifically, an Epanalepsis5 and an Anadiplosis. Another repetition in Example 1 is formed by the word “for”, a Mesodiplosis. Each of these three plokes triggers a subclass of Relative Repetition, i.e., Outer-Boundary Repetition, Cross-Boundary Repetition and Constituent-Medial Repetition. Figure 7 demonstrates the Cross-Boundary Repetition as an example.
The form of an Anadiplosis triggers the neurocognitive afinity of Cross-Boundary Repetition, a subclass of Relative Repetition. Let there be an arbitrary instance c-b-rep1 of Cross-Boundary Repetition that has Elements rep1e1 and rep1e2, each having its own instance of Position, i.e., 4A position can be an absolute position (e.g., position 0, 1, etc.) or a respective or relative position with respect to another element, e.g., immediately after a word, beginning of a clause, etc. For simplicity, position refers to the latter unless otherwise stated.
5Let’s treat . . . , . . . as epanalepsis for now, although it refers to the repetition at the beginning and the end of the same clause.
e1
Refers
To "all" Repetition
rep Before Position
Refers
To position(rep1e1) and position(rep1e2). The elements of c-b-rep1 not only specify the order of their positions with respect to each other, but also with respect to the clauses. A Clause is a Linguistic Element that Contains Token. A Clause is connected to a Relative Position which is a subclass of Position that represents a position within a clause. Other subclasses of Position are Unpatterned Position and Respective Position. A Relative Position can be either Initial, Medial, or Final. A Clause Precedes or Follows another Clause. A Clause can also be Proximal to another Clause. Graphically, the classes Clause and Position can be represented as in Figure 8. The positions of elements of c-b-rep1 must be instances of Final and Initial of the clauses in which the anadiplosis is contained in. For example, let c1 and c2 be instances of Clause and Precedes(c1, c2). The positions of elements rep1e1 and rep1e2 are located in c1 and c2 respectively. In terms of Example 1, the string values of c1 and c2 are “All for one” and “one for all” respectively. Both rep1e1 and rep1e2 refer to the word “one”. In this case, the position of rep1e1 must be an end position of c1. Thus position(rep1e1) is an instance of End that is linked to c1. Similarly, position(rep1e2) is an instance of Initial that is linked to c2. Another relation between the clauses is the Proximal(c1, c2) to indicate proximity between the clauses. Similarly, a Constituent-Initial, Constituent-Medial, and Constituent-Final Repetition triggered by Epanaphora, Mesodiplosis, and Epiphora respectively, has both elements with positions that are instances of Initial, Medial, and Final respectively. In the case of Epanalipsis, positions of elements are the opposite of that in the case of Anadiplosis, i.e., rep1e1 and rep1e2 are instances of Initial and Final respectively.
Another subclass of Repetition is Nested Repetition, in which its elements are themselves instances of Repetition. In other words, a Nested Repetition is the repetition of Repetitions. This paper demonstrates Reversed Repetition, a subclasses of Nested Repetition while also a subclass of Respective Repetition, triggered by the form of an Antimetabole.
The form of an antimetabole triggers a Reversed Repetition, which is a subclass of Repetition. The elements of Reversed Repetition are Repetitions. Let rep1 and rep2 be instances Relative Repetition Before Position Precedes Proximal Anadiplosis
Form rep1e2 position(rep1e2)
Initial
Ploke of Repetition and rep1e1, rep1e2, rep2e1 and rep2e2 be elements of rep1 and rep2 defined as in Figure 6 accordingly, then an instance revRep1 of Reversed Repetition can be defined as follows and graphically in Figure 9:
ReversedRepetition(revRep1) ∧ ElementOf(rep1, revRep1) ∧ ElementOf(rep2, revRep1) → Before(position(rep1e1), position(rep2e1)) ∧ Before(position(rep2e1), position(rep2e2)) ∧ Before(position(rep2e2), position(rep1e2)) In Example 1, rep1e1, rep1e2 both refer to the word “all” and rep2e1, rep2e2 both refer to the word “one”. The order of positions of the elements is, in conventional notation, 11 < 21 < 22 < 12. Note that in this case the positions are not instances of Relative Position, but of Respective Position with relations Before and After representing the order of the position of each element with respect to each other. Since relations Before and After are transitive, the relation Before(position(rep1e1), position(rep1e2)) defined as part of the definition (i.e., the dashed line in Figure 9) of rep1 is implicit from the definition of revRep1.
What about the other neurocognitive afinities such as balance, opposition, etc.? These neu
RefersTo
Token
Position Unpatterned
Position
Respective
Position
Relative
Position Initial
Medial
Final rocognitive afinities are the results of multiple rhetorical figures working together [ 26]. In the scope of Ploke, they are certainly contributed by the subclasses of Repetition. A Clause-Medial Repetition contributes to balance, a Clause-Boundary Repetition contributes to opposition, a Reversed Repetition together with a Clause-Medial Repetition contribute to both balance and opposition, etc. In the case of Example 1, the combination of antimetabole and mesodiplosis produces a stronger balance. Thus, the forms of plokes do not trigger balance and opposition directly, but contribute through combinations of Repetition and Position and their subclasses. That is, repetition and position are the primary neurocognitive afinities triggered by the form of a ploke.
We have proposed an ontology for ploke which treats ploke as a class of figures rather than an isolated figure. The ontology outlines the subclasses of Ploke which includes rhetorical figures of lexical repetition such as Anadiplosis, Epanaphora, Antimetabole, etc. The ontology also models the path of the attentional efect a rhetorical figure has on a receiver. That is, Rhetorical Figure → Form → Neurocognitive Afinity → Attention Efect as in Figure 1.
Another important contribution of the proposed ontology is the classification and representation of repetition and position, both are primary neurocognitive afinities triggered by the form of a ploke. In summary, Anadiplosis triggers the Cross-Boundary Repetition, a subclass of Repetition in which positions of elements are specified as instances of Relative Position, a subclass of Position. Similarly, Epanaphora, Mesodiplosis and Epiphora trigger specific types of repetitions in which positions of the elements, e.g., position(rep1e1) and position(rep1e2), are instances of classes Initial, Medial, and Final respectively, all of which are subclasses of Relative Position. As demonstrated by the Antimetabole in Figure 9, reversal is the alternation of positions of Elements that is part of Reversed Repetition, a subclass of Nested Repetition, in which the Elements are Repetitions.
All neurocognitive afinities triggered by forms of plokes harmoniously generate attentional
efects such as salience, aesthetic pleasure, and increase memorability. Therefore it is crucial to
have a formal knowledge representation model for ploke that includes concepts of
neurocognitive afinities in order for AI/NLP systems to take advantage of rhetorical information embedded
in rhetorical figures. The next step is to extend and construct ontologies for other rhetorical
ifgures, including other schemes, phonological and morphological figures, in similar manner. It
is also possible to incorporate the proposed ontology with existing rhetorical figure detection
tools [
Another possible direction of research is to explore the connection between neurocognitive afinities and the function of a rhetorical figure. We have already expressed a number of functions of ploke in First Order Logic. For example, in Figure 9, the proposed ontology is able to infer that an arbitrary relation exists between the elements rep1e1 and rep2e1 of the Reversed Repetition, triggered by the form of an Antimetabole, whose function is to convey a reciprocal relation. However, contextual information is required to instantiate . Because functions of rhetorical figures are loosely researched in the field of rhetoric, it would greatly benefit the field if the connection between neurocognitive afinities and functions can be established formally. New discoveries of functions will return the favor to Computational Rhetoric and AI/NLP with potential rhetorical figure ontologies that are more comprehensive and robust. We also proposed that repetition and position are the primary neurocognitive afinities that contribute to other compound afinities such as balance and opposition. We are interested in the formal representation of how each variation of repetition contributes to the compound neurocognitive afinities. Therefore, formalizing the connections between these neurocognitive afinities is another research direction in future. [18] D. D. Hromada, Initial experiments with multilingual extraction of rhetoric figures by means of PERL-compatible regular expressions, in: Proceedings of the Second Student Research Workshop, associated with RANLP 2011, 2011, pp. 85–90. [19] M. Dubremetz, J. Nivre, Rhetorical figure detection: The case of chiasmus, in: Proceedings of the Fourth Workshop on Computational Linguistics for Literature, 2015, pp. 23–31. [20] M. Dubremetz, J. Nivre, Machine learning for rhetorical figure detection: More chiasmus with less annotation, in: Proceedings of the 21st Nordic Conference on Computational Linguistics, 2017, pp. 37–45. [21] J. Lawrence, J. Visser, C. Reed, Harnessing rhetorical figures for argument mining, Argument & Computation 8 (2017) 289–310. [22] E. L. Clarke, The Relo-KT Process for Cross-Disciplinary Knowledge Transfer, Ph.D. thesis,
Trinity College Dublin, Ireland, 2019. [23] A. Dumas, The Three Musketeers, volume 1, Sovereign via PublishDrive, 2013. [24] R. A. Harris, Re-inventing rhetorical figures: Celebrating the past, building the future, 2018.
URL: http://www.arts.uwaterloo.ca/~raha/cv/documents/RSA-2018-Harris-Randy.pdf. [25] J. Fahnestock, Rhetorical Figures in Science, Oxford University Press on Demand, 2002. [26] R. A. Harris, Chiastic iconicity, Iconicity in Language and Literature 18 (to appear). [27] J. Fahnestock, Rhetorical Style: The Uses of Language in Persuasion, Oxford University
Press, 2011. [28] R. A. Harris, C. Di Marco, Rhetorical figures, arguments, computation, Argument &
Computation 8 (2017) 211–231. [29] J. F. Kennedy, Inaugural address, January 20, 1961. [30] N. Armstrong, One small step, transcript of Apollo 11 moon landing, July 20, 1969. [31] B. Hutchinson, BC man survives violent encounter with RCMP, National Post (2015). Sect.
A:15, A17. [32] J. Lapidos, The Hottest Rhetorical Device of Campaign ’08, Slate, Inc., 2008. URL: https: //slate.com/human-interest/2008/09/the-hottest-rhetorical-device-of-campaign-2008. html. [33] R. A. Lanham, A Handlist of Rhetorical Terms, University of California Press, Berkeley,
CA, USA, 1991. [34] I. Newton, N. Chittenden, Newton’s Principia: The Mathematical Principles of Natural
Philosophy, Geo. P. Putnam, 1850. [35] M. Johnson, The Body in the Mind: The Bodily Basis of Meaning, Imagination and Reason, The Univ. of Chicago Press, 1987.