See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/280246866

Schemas for Narrative Generation Mined from Existing Descriptions of Plot

Conference Paper · May 2015

DOI: 10.4230/OASIcs.CMN.2015.54

CITATIONS READS

5 5,452

3 authors:

Pablo Gervás Carlos León

Complutense University of Madrid Complutense University of Madrid
250 PUBLICATIONS 2,768 CITATIONS 69 PUBLICATIONS 397 CITATIONS

SEE PROFILE SEE PROFILE

Gonzalo Méndez
Complutense University of Madrid
80 PUBLICATIONS 519 CITATIONS

SEE PROFILE

All content following this page was uploaded by Gonzalo Méndez on 22 July 2015.

The user has requested enhancement of the downloaded file.

Schemas for Narrative Generation Mined from
Existing Descriptions of Plot∗
Pablo Gervás1 , Carlos León2 , and Gonzalo Méndez3

1 Instituto de Tecnología del Conocimiento, Universidad Complutense de Madrid

Ciudad Universitaria, 28040 Madrid, Spain
[email protected]
2 Facultad de Informática, Universidad Complutense de Madrid
Ciudad Universitaria, 28040 Madrid, Spain
[email protected]
3 Facultad de Informática, Universidad Complutense de Madrid
Ciudad Universitaria, 28040 Madrid, Spain
[email protected]

Abstract
Computational generation of literary artifacts very often resorts to template-like schemas that
can be instantiated into complex structures. With this view in mind, the present paper reviews a
number of existing attempts to provide an elementary set of patterns for basic plots. An attempt
is made to formulate these descriptions of possible plots in terms of character functions, an
abstraction of plot-bearing elements of a story originally formulated by Vladimir Propp. These
character functions act as the building blocks of the Propper system, an existing framework for
computational story generation. The paper explores the set of extensions required to the original
set of character functions to allow for a basic representation of the analysed schemata, and a
solution for automatic generation of stories based on this formulation of the narrative schemas.
This solution uncovers important insights on the relative expressive power of the representation
of narrative in terms of character functions, and their impact on the generative potential of the
framework is discussed.

1998 ACM Subject Classification F.4.1 Knowledge Representation Formalisms and Methods

Keywords and phrases Narrative generation, conceptual representation of narrative, character

functions, plot, narrative schemas

Digital Object Identifier 10.4230/OASIcs.xxx.yyy.p

1 Introduction
Computational generation of literary artifacts very often resorts to template-like schemas
that can be instantiated into complex structures. This approach has been addressed in the
story generation field as a number of computational systems following a grammar-based
design [9, 6, 5].
With this view in mind, the present paper reviews a number of existing attempts to
provide an elementary set of patterns for basic plots. None of these attempts have been
accepted as generally valid. To a large extent, they rely on oversimplification – reducing plot
to a very abstract outline that conforms to a great number of story but characterises none of
them –, or they focus on particular aspects of a given story – to the detriment of others – so

∗
This work was partially supported by FP7 WHIM project Grant Agreement 611560.
© Pablo Gervás, Carlos León and Gonzalo Méndez;
licensed under Creative Commons License CC-BY
Computational Models of Narrative.
Editor: Mark Finlayson Editor; pp. 1–17
OpenAccess Series in Informatics
Schloss Dagstuhl – Leibniz-Zentrum für Informatik, Dagstuhl Publishing, Germany
2 Schemas for Narrative from Existing Descriptions of Plot

it can be reduced to a schema that matches a larger number of stories. Such characteristics
may play against the usefulness of any particular one of them as single framework for the
description or classification of stories. However, considered as a whole, they can be understood
as a basic abstract vocabulary to describe different plots. In the context of automated story
generation, such a vocabulary would be very useful in at least two different senses:
it may provide an agreed vocabulary for describing what type of story is desired, e.g. “a
vengeance story” or “a quest story”
it may provide a basic skeleton that the desired story should satisfy, regardless of any
additional complexity that may be introduced to enrich it

In order to address needs of this kind, the present paper attempts to formulate these
descriptions of possible plots in terms of schemas that may be used to drive the Propper
system, an existing framework for computational story generation. The paper also explores
the set of extensions required to the original set of character functions to allow for a basic
representation of the analysed schemata. This is intended as a proof of concept to test the
initial hypothesis of the usefulness of such schemas in the context of story generation. The
Propper system [3, 4] is a computational implementation of the procedure for generating
stories described by Vladimir Propp [8] as a possible use of his classic formalization of the
morphology of the folk tale.
Once the various descriptions for plot are available as schemas that can be used to drive
the Propper system, the impact of using them instead of - or as well as - the original canonical
sequence for folk tales is discussed in terms of whether it expands the generative potential of
the Propper system.

2 Review of Previous Work

This section reviews some of the existing proposals for the schematisation of possible story
plots, the Proppian morphology of a folk tale, and the Propper system for story generation.
Later sections bring these ingredients together to propose a computational model of narrative
that can consider input in terms of the reviewed plot schemas and produces matching stories.

2.1 Some Existing Descriptions of Schemas for Plot

Christopher Booker [2] proposes that there are seven basic plots such that all possible stories
can be seen as instantiations of these. The seven plot in question are described briefly in
Table 1. These descriptions attempt to capture the basic outline for purposes of reference,
more detailed descriptions follow below.
An important point to note is that these plots are not mutually exclusive. Any given
narrative may combine several of them into its overall structure, with some of these subplots
possibly focusing on different characters.
Tobias [10] proposes the existence of 20 master plots. His book is more oriented towards
instruction on how to build instances of these plots. A relevant insight presented here is that
plots can be divided into plots of the body – involving mainly action – and plots of the mind
– involving psychological development of the characters. Brief descriptions of these 20 master
plots are provided for reference in Table 2.
The 20 plots by Tobias are even more difficult to keep separate from one another in
practical terms. In terms of actual events in the narrative, quests or adventures are very
likely to include elements of pursuit, rescue, escape, rivalry, revenge, temptation, sacrifice,
or some character being an underdog at some stage. In terms of character development,
P. Gervás, C. León and G. Méndez 3

Table 1 The Seven Basic Plots as described by Booker

Overcoming the Monster hero sets out to confront a monster and eventually defeats it
Rags to Riches hero starts from humble beginnings and eventually achieves
happiness
The Quest hero sets out to fulfill a quest
Voyage and Return hero sets out on a journey and returns having matured in the
process
Comedy initial confusion involving love relationships is eventually re-
solved happily
Tragedy traces the fall from grace of a particular character to a tragic
ending
Rebirth main character almost falls from grace but repents at the last
minute

Table 2 20 Master Plots as presented by Tobias

Quest hero sets out to fulfill a quest

Adventure much like a Quest but with less focus on a particular goal and
more action
Pursuit hero is pursued and eventually manages to escape
Rescue hero rescues a victim imprisoned by a villain
Escape like Rescue but the protagonist is the victim and eventually
escapes by his own means
Revenge protagonist sets out to avenge a villainy
The Riddle involves solving a riddle (reader should try to solve it before the
protagonist)
Rivalry a protagonist and an antagonist of balanced power clash, prot-
agonist wins
Underdog as in Rivalry but protagonist is at disadvantage and wins through
tenacity
Temptation maps the fight of protagonist against temptation, from initial
fall to eventual success
Metamorphosis protagonist suffers a curse that transforms him into a beast, but
love releases him eventually
Transformation faced with a crisis, protagonist suffers transformation with im-
portant effects (usually at a price)
Maturation tracks immature character through challenging incidents to
maturity (usually achieved at a price)
Love maps the progress of a love relation from initial obstacles to
final fulfillment (if test passed)
Forbidden Love as in Love but around an unconventional love relation (usually
adultery) which ends badly
Sacrifice tracks transformation of main character from low to high moral
state, leading to a final sacrifice
Discovery protagonist discovers himself
Wretched Excess traces psychological decline of a character based on a character
flaw
Ascension protagonist faces a moral dilemma and undergoes ups and down
till he reaches success
Descension as in Ascension but followed to final disaster

CMN 2015
4 Schemas for Narrative from Existing Descriptions of Plot

Table 3 The 36 dramatic situations as described by Polti.

Supplication power in authority must choose between a persecutor

and a suppliant
Deliverance protector comes to the rescue of the distressed
Crime Pursued by Vengeance avenger executes a vengeance on a criminal
Vengeance taken for kindred avenger and the criminal are kin
upon kindred
Pursuit hero is pursued by an abstract peril or punishment
Disaster a power is defeated by an enemy or catastrophe
Falling Prey to Cruelty of Mis- hero suffers a cruel master or misfortune
fortune
Revolt hero is a conspirator that intrigues against a tyrant
Daring Enterprise hero attempts to recover an object or person from an
adversary
Abduction hero rescues an abducted victim from its abductor
The Enigma a combat of the intelligence to find a person or object
Obtaining aim to be achieved through eloquence and diplomacy
Enmity of Kinsmen kinsmen transform love into (usually) mutual hatred
Rivalry of Kinsmen a desired person causes a kinsman to hate another
Murderous Adultery a betrayed husband or wife kills one or both adulterers
Madness a madman slays, injures or brings disgrace onto a victim
Fatal Imprudence imprudence or curiosity as the cause of a loss
Involuntary Crimes of Love character unknowingly commits adultery or incest
Slaying of a Kinsman Unrecog- unrecognized victim is slain by a kinsman
nized
Self-Sacrifice for an Ideal hero sacrifices life, love or well-being to a cause
Self-Sacrifice for Kindred hero makes sacrifices for happiness of a relative
All Sacrificed for Passion character makes sacrifices for a vice or passion
Necessity of Sacrificing Loved hero sacrifices a loved one for a necessity or vow
Ones
Rivalry of Superior and Inferior two masculine or feminine rivals with different rank
Adultery a deceived husband or wife
Crimes of Love a lover and beloved incur in questionable acts
Discovery of the Dishonor of a a character discovers the shame of a loved one
Loved One
Obstacles to Love marriage prevented by social norms
An Enemy Loved one of two lovers is hated by kinsmen of the other
Ambition character tries to obtain a good guarded by an ad-
versary
Conflict with a God a mortal struggles with a deity
Mistaken Jealousy a character is jealous of another
Erroneous Judgement any kind of mistaken judgement
Remorse a culprit suffers remorse for a crime or love fault
Recovery of a Lost One a hero struggles to find a lost, loved one
Loss of Loved Ones a character witnesses the death of a loved one

they may also include transformation, maturation, or discovery. Much the same may be said
about love stories. Our understanding it that a plot is considered to satisfy one of these
labels only if the label is applicable to the main structure of the plot.

Georges Polti [7] proposed 36 dramatic situations, following Gozzi’s assertion that there
can only be thirty six tragic situations. These situations are briefly described for reference in
Table 3, although Polti divides each of them into a series of classes and sub-classes that are
further described or exemplified in the referenced book.

These 36 situations can be combined in the same story, since they must be understood as
an outcome of previous events in the story, when the intervening characters come together
and the main character in the situation must face a decision to be made, a change to be
suffered or an obstacle to be overcome.
P. Gervás, C. León and G. Méndez 5

2.2 Proppian Morphology of a Story

At the start of the 20th century, Vladimir Propp [8] identified a set of regularities in a subset
of the corpus of Russian folk tales collected by Afanasiev [1]. These regularities he formulated
in terms of character functions, understood as acts of the character, defined from the point
of view of their significance for the course of the action. Character functions are so named
because, in Propp’s understanding, they represent a certain contribution to the development
of the narrative by a given character. According to Propp, for the given set of tales, the
number of such functions was limited, the sequence of functions was always identical, and all
these fairy tales could be considered instances of a single structure.
The set of character functions includes a number of elements that account for a journey,
a number of elements that detail the involvement of the villain – including the villainy itself,
some possible elaborations on the struggle between hero and villain, and a resolution –, a
number of elements that describe the dispatching of the hero, a number of elements that
describe the acquisition of a magical agent by the hero, and a number of elements concerned
with the progressive unveiling of the hero’s role in opposition to a false hero.
It is less well known that Propp provides in his book a very clear description of how his
morphology could be used for story generation.

2.3 The Propper System

The Propper system developed by Gervás [3] constitutes a computational implementation
of a story generator initially based on Propp’s description of how his morphology might be
used to generate stories.
It relies on the following specific representations for the concepts involved:
a character function, a label for a particular type of acts involving certain named roles
for the characters in the story, defined from the point of view of their significance for the
course of the action
a sequence of character functions chosen as backbone for a given story
possible instantiations of a character function in terms of specific story actions, involving
a number of predicates describing events with the use of variables that represent the set
of characters involved in the action

Based on these representations the Propper system defines a procedure that first chooses
a sequence of character functions to act as abstract narrative structure to drive the process,
and then progressively selects instantiations of these character functions in terms of story
actions to produce a conceptual representation – in terms of an ordered sequence of predicates
– of a valid story.
To fulfill Propp’s description of the morphology of a folk tale, the sequence of character
functions that acts as backbone for a story has to be a subset of the character functions
listed by Propp, appearing in a relative order that conforms with a given canonical sequence.
The actual set of character functions employed as canonical sequence is given in Table 4.
Character functions are presented in two columns by their abbreviated name. A key point
in the canonical sequence is the villainy / lack pair of character functions written in bold.
These differ from all the others in that only one of them is ever included in any single story,
and all stories must contain either one or the other.
From a given sequence of character functions, the system defines a fabula, a sequence
of states that contain a chain of story actions – which are instances of those character
functions. A story action involves a set of preconditions – predicates that must be present
in the context for continuity to exist –, and a set of postconditions – predicates that will

CMN 2015
6 Schemas for Narrative from Existing Descriptions of Plot

Table 4 Set of character functions employed as canonical sequence.

test by donor difficult task

hero reaction branding
acquisition magical agent victory
villainy / lack task resolved
hero dispatched trigger resolved
begin counteraction return
acquisition magical agent hero pursued
departure rescue from pursuit
test by donor unrecognised arrival
hero reaction unfounded claims
acquisition magical agent false hero exposed
transfer transfiguration
trigger resolved branding
unrecognised arrival villain punished
unfounded claims hero marries
struggle

be used to extend the context if the action is added to it. Each story action is linked to its
context of occurrence by having its preconditions satisfied by the preceding state. The initial
state by default incorporates all predicates of the first action, and each valid action added to
the fabula generates a new state that incorporates all predicates of the previous state, plus
the predicates of the new action. To evaluate whether the preconditions of a story action are
satisfied by the context, they are unified with the set of predicates that hold in that state.
The revised version described in [4] describes extensions to the original constructive
procedure that take into account the possibility of dependencies between character functions
– such as for instance, a kidnapping having to be resolved by the release of the victim – and
the need for the last character function in the sequence for a story to be a valid ending for it.

3 Describing Existing Schemas for Plots in Terms of Proppian

Character Functions

We want to attempt to unify the material reviewed in section 2 into a single representation
that is compatible with the existing framework of the Propper system. As the Propper
system is driven by Proppian character functions, we will consider whether the schemas
arising from the approaches reviewed can be described as sequences of character functions as
described by Propp, and what extensions might be required for a better fit.

3.1 Establishing a Common Vocabulary from the Set of Taxonomies

The different sets of plots reviewed in section 2.1 show a certain overlap in some cases (both
Booker and Tobias include a plot based on a quest, for instance). Where they differ, it would
be ideal to establish some way in which the elements in one set might be related to elements
in the other, either as more specialised or more abstract versions.
When trying to cross-relate these various taxonomies with one another, it becomes
apparent that they are formulated at different levels of abstraction, and focused on different
aspects of the plot. This makes it difficult to find a clear correlation between them. However,
for the purposes of our paper – which aims at making it possible to rely on these descriptions
to specify desired stories and/or drive the process of their construction – it becomes important
to be able to understand how elements from these descriptions might combine or interact.
In that sense, a number of patterns can be identified. Tobias’ and Booker’s plots can be
related as follows:
P. Gervás, C. León and G. Méndez 7

Tobias’ plots of Temptation, Metamorphosis, Transformation, Maturation and Discovery

could fit Booker’s description of Rebirth plots.
Tobias’ plots of Pursuit, Rescue, Escape, Rivalry, Underdog, Revenge, Sacrifice might be
employed to articulate what Booker describes as an Overcoming the Monster plot.
Tobias’ Love plot correlates nicely with Booker’s Comedy plot.
Tobias’ plots of Wretched Excess, Descension, Forbidden Love, and, possibly, Sacrifice
might fit Booker’s Tragedy plot.
Tobias plot of Ascension fits Booker’s Rags to Riches plot.
Tobias’ plots of Transformation, Maturation and Discovery could apply as descriptions
of character development implicit in Booker’s description of Quest, Voyage and Return,
Rags to Riches and Rebirth plots.

Polti’s dramatic situations are not presented as candidates for complete plots, but rather
as situations with dramatic potential that may arise within a given plot. In this sense, they
are easier to place with respect to the other two proposals considered in this paper. In a
sense, they constitute a finer grained vocabulary for describing plot elements that may occur
in larger plot structures. For this reason, some of them show a surprising match with those
plots of Tobias’ that we have described as elements sometimes used as ingredients being
expanded into full independent plots, such as Pursuit – which appears in both Tobias’ and
Polti’s lists –, or Deliverance in Polti closely matching Rescue in Tobias.
For this set of situations, the task to be considered becomes more to identify where in
the more elaborate structures these situations appear.

3.1.1 Paraphrasing Plot Options in Terms of Character Functions

Booker’s set of seven plots can be easily paraphrased in terms of Proppian character functions.
One such paraphrase of them is given in Table 5. There are some differences. Where Propp
considers a fixed sequence of character functions from which a selection can be picked out,
Booker’s descriptions differ in at least two ways. First, they sometimes allow for more
than one possible relative ordering between some of the elements included. In the table,
this has been represented by placing between brackets those elements that may occur in
interchangeable order or that are optional. Second, Booker’s descriptions include a certain
possibility of some subsequences reoccurring repeatedly over the same plot. In the table,
such subsequences have been replaced with labels in capital letters that have been defined
separately. It may pay to abstract them into higher order labels that can appear within more
structured sequences. They correspond to:

MONSTERS: struggle, hero pursued, (victory, rescue from pursuit)

TESTERS: test by donor, hero reaction, acquisition magical agent

Where certain character functions (or labels for subsequences) can occur more than once
according to Booker, these have been marked with an asterisk *. The case of Tragedy and
Rebirth is strikingly different. Both can indeed be phrased in terms of Proppian character
functions as shown in the table. However, this requires a slight revision of the Proppian
concept of character function. Proppian character functions assume a fixed set of roles,
namely a hero, a villain and some auxiliary characters such as dispatcher, a donor, a helper...
But in Proppian functions, the protagonist of the story is assumed to be always the hero.
In the case of Booker’s Tragedy and Rebirth, the paraphrase works only if the protagonist
is considered to be the villain. This implies that the Tragedy plot would correspond to an
instance of the Overcoming the Monster plot but told from the point of view of the villain.

CMN 2015
8 Schemas for Narrative from Existing Descriptions of Plot

Table 5 Paraphrases of Booker’s 7 basic plots in terms of Proppian character functions

Overcoming the Monster (villainy*, MONSTERS*), struggle, victory, villain punished,

hero marries
Rags to Riches lack, departure, transfiguration, hero marries
The Quest (hero dispatched, difficult task), departure, (MONSTERS*,
HELPER*), task resolved
Voyage and Return departure, ((difficult task, task resolved), (MONSTERS*,
HELPER*)), return
Comedy lack, (transfiguration, unrecognised arrival), (difficult task, task
resolved)*, (hero recognised), transfiguration, hero marries
Tragedy (villainy*, MONSTERS*), struggle, victory, villain punished
Rebirth (villainy*, MONSTERS*), repentance, repentance rewarded

It is important to note that the occurrence of the victory character function now implies that
the protagonist is defeated, which is contrary to Propp’s original interpretation. The Rebirth
plot requires a more elaborate reworking to be phrased in terms of Proppian functions,
because it involves a particular turn in the story that was not originally contemplated by
Propp. This is the point in the narrative where the villain sees the light, repents, and redeems
himself. New character functions would need to be introduced to cover this process, as it
plays a fundamental role in such stories that would definitely need capturing. We refer to
these character functions as repentance and repentance rewarded, and we include them as
such in the table.
The Comedy plot requires a special analysis. It may be phrased in terms of Proppian
functions, in as much as it starts from an initial lack – though specifically related to love, lack
of a love partner, lack of attention from the chosen partner, or lack of permission to marry
the chosen partner –, it involves solving a difficult task – related to the corresponding lack –,
and it ends with the hero marrying. However, the description of this plot provided by Booker
addresses the corresponding story at a level of detail that cannot be covered appropriately
with Proppian functions, at least in the sense that these had been defined within the Propper
system. To deal with this case, we would need a system with the following features:
the ability to explicitly represent the gender of characters1 , as the core of the plot revolves
around love relations between characters
the ability to represent shifts in affinity between characters and to have these shifts arising
from and triggering events in the narrative
the ability to consider a number of interwoven subplots focused on different characters

Such features are beyond the scope of the present paper but they will be considered for
future work. Nevertheless, a basic sketch of the Comedy plot in terms of Proppian functions
has been provided for completeness.
According to Booker’s description of his plots, the paraphrases given in Table 5 constitute
a sketch of the main events that characterise each of the plots. The fleshing out of these
plots into specific actual stories may involve combining more than one plot, in which case
the corresponding sequences of character functions may intermingle as different narrative
threads. When such task is attempted computationally, some means must be provided for
keeping track of which characters play which roles in which of these threads, and whether

1
Although in current times it might have been more politically correct to phrase this in terms of sexual
preferences, we have opted in this desiderata for a more classical approach to character pairings in terms
of gender. This might be revised in future work to allow for more generic and politically correct story
telling capabilities.
P. Gervás, C. León and G. Méndez 9

Table 6 Paraphrases of the Elementary Plots of Tobias’ in terms of Proppian character functions

Pursuit hero pursued, rescue from pursuit

Rescue villainy, trigger resolved
Escape villainy, trigger resolved [protagonist is victim, not hero!]
Revenge villainy, villain punished
The Riddle difficult task, task resolved
Rivalry struggle, victory
Underdog struggle, victory [protagonist at disadvantage]

any given character can play different roles in different threads. This is beyond the scope of
the present paper and it is currently left for future work.
As discussed in section 3.1, the elements described by Tobias amongst his 20 master
plots operate at a slightly different level of abstraction from those used by Booker. In a
certain sense, they correspond to focusing the plot of a complete story on particular types of
situation that were occurring as parts of the plots considered previously. The correspondences
already established between Booker’s and Tobias’ plots introduce a change in the overall
task definition. Given that many of the plot descriptions given by Tobias can be seen as
specific instances of Booker’s plots, it is less useful to paraphrase them in terms of Proppian
functions – the paraphrase already given for the corresponding Booker plot might be used in
each case – and it becomes more interesting to consider how the different instantiations that
Tobias provides might be differentiated from one another in terms of a Proppian description
(or what extensions of the Proppian implementation might be required to consider these
plots).
Tobias’ plots of Pursuit, Rescue, Escape, Rivalry, Underdog, Revenge, Sacrifice can be
represented as more specific plots that focus on parts of the sequences of character functions
used to describe Booker’s plots. A tentative paraphrasing for them is presented in Table 6.
The Quest and Adventure plots can be seen as similar to Booker’s The Quest and Voyage
and Return. Tobias’ Love plot has been linked to Booker’s Comedy plot, and so it is subject
to the same considerations described earlier for that one. The Ascension plot can be mapped
to the Rags to Riches plot.
The remaining plots described by Tobias can be grouped into a set of instantiations of the
two Booker plots already described that presented significant differences with the Proppian
schema: Tragedy and Rebirth.
Forbidden Love is related to Comedy/Love plots in that its main ingredient is a love
relationship, and it differs from them in two ways: the love relation in question is one against
convention, and it ends badly. As before, this may be implemented using the same set of
characters and actions as for comedy, but making the protagonists a pair of characters that
do not get paired off in the end. This is similar to the opposition between Overcoming the
Monster and Tragedy. In a sense, one could say that Tobias is enriching the set of plots by
considering a plot based on love but which can end badly, whereas Booker only considers
plots on love that end well.
In a similar opposition, the Descension and Wretched Excess plots could be seen as dark
counterparts to the Rags to Riches/Ascension type of plot. These may be paraphrased in
terms of Proppian functions by inverting the order in which the functions in the sequence for
Rags to Riches occur. However, better results might be obtained if specific character functions
are defined to represent: an initial positive situation for the character – corresponding to
a positive version of lack –, a character function to discover events in which the fortune of
the protagonist suffers, and a final negative situation. This suggests that a reworking of the
set of character functions might benefit from a little generalization, so that both positive

CMN 2015
10 Schemas for Narrative from Existing Descriptions of Plot

and negative situations can be described, and events that cause transitions in both positive
and negative directions can be represented. Then the opposing pairs of plots may all be
represented based on these. The original set of character functions defined by Propp covers
only part of this spectrum – it includes no character function for a positive initial situation –
and relies on very specific solutions for some particular areas – it links very tightly the final
positive situation of the hero with either marriage or coronation, for instance. An effort to
broaden this set of character functions would greatly improve the range of possible stories
that can be generated. As this requires a heavy effort of knowledge engineering of system
resources it is postponed for future work.

Differences between Descension and Wretched Excess can be identified in terms of one
being more concerned with material situation of the protagonist, and the other with his/her
psychological decline. In marking this difference, Tobias shows a concern with an aspect
of plots that had not been considered by either Propp or Booker: the difference between
physical and psychological characterization.

The set of plots proposed by Tobias shows an increase in number partly because it distin-
guishes a number of plots that are based on psychological development of their protagonists –
what he describes as plots of the mind – beyond those considered by Propp – which centre
almost exclusively on what Tobias calls plots of the body. These plots of the mind are the
Temptation, Transformation, Maturation and Discovery plots. The Metamorphosis plot
combines such a psychological ingredient with a physical change. In terms of Booker’s classi-
fication, most of these qualify as Rebirth plots, as they involve a change of the protagonist
during the development of the plot. In a certain sense, the Sacrifice plot also includes a
similar turning point related to psychological issues, though in this case the change also
translates into a physical sacrifice. The differences between the various plots arise from these
slight differences in the relative importance of the material and the psychological aspects, or
in the specific type of change that the protagonist is subjected to – as described reasonably
well by the names of these plots.

Again, the representation of the psychological evolution of characters is beyond the

current capabilities of the Propper system, and discussion of an appropriate extension beyond
the scope of the present paper, but it will be considered as future work.

With respect to Polti’s dramatic situations, these are not so much patterns for complete
plots but rather building blocks that may be employed in the construction of plots. In this
sense, they are closer to being descriptions of actions of the characters that are significant for
the course of the action, which is what Propp’s character functions are intended to be. For
this reason, when establishing a correspondence that might lead to a common vocabulary for
plot descriptions, it would be more useful to consider Polti’s dramatic situations as alternative
abstractions, closely related to Proppian character functions. A possible alignment between
Polti’s dramatic situations and Propp’s character functions (or groups thereof) is shown
in Table 7. The material is presented according to the following criteria. For each line of
the table, the first column indicates a character function or a group of character functions
that might be considered to correlate in some way with the dramatic situations listed in
the second column. The third column is used to indicate specific characteristics that the
instantiations of the character functions given in the first column would need to satisfy to
properly represent the dramatic situation given in the second column. The bottom half of the
table shows dramatic situations that have no direct match to Proppian character functions.
For these, it may be worth considering the introduction of specific character functions.
P. Gervás, C. León and G. Méndez 11

Table 7 Alligment of Polti’s 36 Dramatic Situations with Proppian character functions

lack Ambition
Recovery of a Lost One
Loss of Loved Ones
lack Disaster
villainy Falling Prey to Cruelty of Misfortune
Madness
Fatal Imprudence
Involuntary Crimes of Love
Slaying of a Kinsman Unrecognized
Adultery (love)
Crimes of Love (love)
Discovery of the Dishonor of a Loved One
trigger resolved Deliverance
rescue from pursuit
victory Crime Pursued by Vengeance
villain punished Vengeance taken for kindred upon kindred
trigger resolved
hero pursued Pursuit
struggle Enmity of Kinsmen (psychological)
Rivalry of Kinsmen
Rivalry of Superior and Inferior
trigger resolved Abduction
Murderous Adultery
test by donor Daring Enterprise
hero reaction The Enigma (temptation or a riddle)
acquisition Obtaining
/
difficult task
task resolved
Self-Sacrificing for an Ideal (sacrifice)
Self-Sacrifice for Kindred (sacrifice)
All Sacrificed for Passion (sacrifice)
Necessity of Sacrificing Loved Ones (sacrifice)
Obstacles to Love (love)
An Enemy Loved (love)
Mistaken Jealousy (psychological)
Erroneous Judgement (psychological)
Remorse (psychological)
Supplication
Revolt
Conflict with a God

CMN 2015
12 Schemas for Narrative from Existing Descriptions of Plot

3.2 Extending the Propper System for Schema-Driven Generation

Once a common vocabulary has been agreed that includes elements from the various tax-
onomies, the Propper system has been extended to take advantage of it.
This implies two basic extensions beyond the previous versions of the system:

it must accept input in the form of elements from this vocabulary to drive the story that
is to be constructed
it must be capable of producing stories that match the corresponding description

The first extension has been achieved by means of a preprocessing module that, given the
name of a given narrative schema, builds a sequence of character functions based on resources
along the lines of the tables presented in Section 3.1.1. To build a proof of concept, the
complexities of repetition and alternative ordering have not been considered and the initial
version focuses on simple instantiations of the more generic sequences. These sequences can
now be used as input to the stage of fabula generation of the Propper system, which searches
for appropriate instantiations of these character functions in terms of story actions that link
into a coherent whole that can be recognisable as a story.
The second extension has proven to be more difficult, but it has also uncovered a number
of important insights on the advantages and disadvantages of Propp’s framework as a
computational model of narrative. Additionally, this effort has prompted a number of
improvements that have allowed the system to go beyond Propp’s original formulation.
The first insight relates to the fact that most of the sequences required to implement the
set of narrative schemas reviewed were already included in the canonical sequence proposed
by Propp. This must be considered an important merit of Propp’s framework as it implies
that the method for story generation outlined by Propp – in terms of selecting character
functions from his canonical sequence and instantiating them – would in theory be capable
of producing instances of most of the narrative schemas reviewed. The difficulty would lie in
how to inform the choices at each point. This is part of the problem that the rest of this
section attempts to address.
The second insight concerns the fact that the set of story actions developed to cover the
Proppian character functions includes a broad range of possible story actions to instantiate
each character function. However, in many cases the specific instances of character function
occurring in the context of one of these more specific narrative schemas need to be restricted
to a subset of the complete range of possible story actions. For instance, when the character
function for lack occurs at the beginning of a Rags to Riches schema it works better if
instantiated with story actions concerned with hardship or poverty rather than desire for
wondrous magical objects, whereas both occur in the context of Proppian tales. When
the same character function occurs at the beginning of a Comedy plot, it only works if
instantiated with story actions concerned with lack of a love partner, or lack of permission
to marry. To address this issue, the module of the Propper system concerned with retrieving
possible story actions to instantiate a given character function has been refined to take into
account what particular narrative schema is being considered in each case. The knowledge of
which story actions are suitable to instantiate which character functions under particular
narrative schemas has been encoded explicitly in resources local to these modules. A similar
mechanism may be applied to address the more detailed specific instantiation of character
functions required to generate instances of Tobias’s plots and/or Polti’s dramatic situations,
as described above.
A third important insight arose from the observation that, whereas the Proppian mor-
phology takes for granted that the protagonist of the stories is always the hero, some of the
P. Gervás, C. León and G. Méndez 13

set of narrative schemas considered focused on the villain as protagonist. Namely, Booker’s
schemas for Tragedy and Rebirth, and those of Tobias’s plots that in the analysis in Section
3.1 have been associated to these two. This presents no problem to our endeavour in as much
as the conceptual representation of a story as currently produced by the Propper system is
agnostic as to who is the protagonist. This will become apparent in the examples presented
later in the paper. This issue of who the protagonist is would have to be taken into account
in future work, once the problem of rendering these conceptual representations of stories as
text is addressed.
A fourth insight focused on the fact that to obtain sequences of character functions that
matched as closely as possible the descriptions of the narrative schemas, certain character
functions (or subsequences thereof) might need to occur more than once. This presented
problems because not all instances of the available story actions allowed this. For instance,
some of the story actions for the victory character function allowed the villain to survive
the encounter – thereby being available for a second struggle later in the story –, whereas
others ended more radically with his demise. This restriction was particularly important
to distinguish between the two types of schema where the villain acts as protagonist of the
story: instances of the Rebirth narrative schema require the villain to repent at some point in
the story and undergo a radical change for good, whereas instances of Tragedy may well end
in his utter destruction. From a computational point of view, it required a solution capable
of discerning which particular story actions could be used to instantiate a character function
at what points of the story. The process for selecting story actions was refined further to
take into consideration the relative position of each character function within the narrative
schema being considered.
The observed possibility of repeating and combining certain subsequences of character
functions to make up more complex schemas led to a fifth insight concerning Propp’s
morphology. Although the canonical sequence of character functions as described by Propp
includes a certain redundancy to allow character functions (or small subsequences of them) to
occur at more than one point in the overall narrative arch, the morphology as formalised is too
rigid to capture appropriately the broad range of narrative schemas that have been reviewed.
Propp’s insistence that the character functions in his morphology need be considered in
a specific order introduces a restriction that reduces the expressive power that it might
otherwise have had. This is particularly relevant given that the set of narrative schemas
reviewed is by definition a subset of all the possible ones. For this reason, we intend to
address as future work alternative possible means of combining these sequences of character
functions into complex narrative schemas.

3.3 Examples of Constructed Stories Matching Given Narrative

Schemas
Although it would be impossible to include in this paper examples of stories to match all
the various narrative schemas reviewed, an effort has been made to cover instances of at
least the seven basic plots described by Booker. As the other narratives schemas or dramatic
situations have been related back to these seven in the sections above, this should be seen as
an indication of the potential of the approach.
The task of extending the knowledge resources of the system to cover the full set of
schemas would be significant. The original knowledge engineering effort for the first version of
the Propper system, as reported in [3], demonstrated this task to be an important bottleneck
for the development of this type of system. As a proof of concept, a basic initial version of the
desired approach has been implemented based on the existing resources in terms of related

CMN 2015
14 Schemas for Narrative from Existing Descriptions of Plot

Table 8 An example story for the Overcoming the Monster narrative schema

0 character id810
0 torment_at_night id810 id811
0 victim id811
0 character id811
0 misbehaved id810
1 runs_away id811
1 pursues id810 id811
1 demands id810 id811
2 hides id316 id811
2 escapes id811
3 weight_contest id811 id810
3 confrontation id811 id810
4 heavier id811
5 punished id810
5 shot id810
6 marries id811
6 acceeds_to throne id811

Table 9 An example story for the Rags to Riches narrative schema

0 character id301
0 lack id301 money
1 sets_out id301
2 builds id301 palace
2 new_physical_appearance id301
3 marries id301

sets of character functions and story action resources. The two new character functions
repentance and repentance rewarded and a small set of possible instantations of them as story
actions have been added. The stories that result from this effort are reported below.
Table 8 presents an example of story corresponding to the Overcoming the Monster
narrative schema. This particular story has the peculiarity that the system has picked the
victim of the initial villainy as the hero of the story.
Table 9 presents an example of story corresponding to the Rags to Riches narrative
schema. This story is indicative of how the simplest structure that conforms to one of these
schemas may be insufficient to hold the reader’s interest and fleshing out with additional
narrative elements may be required.
Table 10 presents an example of story corresponding to the Comedy narrative schema.
As indicated above, this is intended only as a baseline. Quality would improve significantly
once the complexities outlined earlier as required for Comedy are addressed.

Table 10 An example story for the Comedy narrative schema

0 character id298
0 lack id298 bride
1 puts_on id298 garment
1 deceiving_appearance id298
2 arrives id298 id719
2 location id719
2 disguised id298
2 unrecognised id298
3 sets id157 id298
3 character id157
3 involves difficult_task hiding
4 solve id298 difficult_task
4 before dead_line
5 recognised id298
6 puts_on id298 garment
6 new_physical_appearance id298
7 betrothed id298
P. Gervás, C. León and G. Méndez 15

Table 11 An example story for the Tragedy narrative schema

0 character id775
0 substitute id775 id776 id777
0 victim id776
0 character id776
0 bad id777
0 misbehaved id775
1 runs_away id776
1 pursues id775 id776
1 demands id775 id776
2 throws id776 id310
2 turns_into id310 id312
2 obstacle id312
2 escapes id776
3 weight_contest id776 id775
3 confrontation id776 id775
4 heavier id776
5 punished id775
5 shot id775

Table 12 An example story for the Rebirth narrative schema

0 character id805
0 try_to_eat id805 id806
0 victim id806
0 character id806
0 misbehaved id805
1 runs_away id806
1 pursues id805 id806
1 demands id805 id806
2 turns_into id806 id314
2 unrecognisable id314
2 escapes id806
3 play id806 id805 cards
3 confrontation id806 id805
4 wins id806
5 repents id805
6 acceeds_to throne id805

Table 11 presents an example of story corresponding to the Tragedy narrative schema. It

is important to note that in this story the protagonist must be considered to be character
id775, who plays the role of the villain.
Table 12 present an example of story corresponding to the Rebirth narrative schema.
Again, the protagonist of this story is character id805.
The stories for narrative schemas corresponding to The Quest and Voyage and Return as
described rely heavily on a combination of a number of incidents. As a result, they turned
out to be overlong to be reported within the size limitations of the paper, but the system
has been extended to be able to produce them. They also suffer from the rigid sequencing of
the various elements involved (struggles with villains, chases, task to solve, encounters with
magical helpers). The more flexible solution for the relative ordering of these elements that
is being considered as future work would result in better stories.

4 Discusion
The extensions that have been required to enable the representation of existing plot schemas
as paraphrases in terms of Proppian character functions arose from one of two possible
situations:
the plots in question violated one of Propp’s basic premises (which basically involve the
protagonist being the hero and the tale having a happy ending)

CMN 2015
16 Schemas for Narrative from Existing Descriptions of Plot

the set of character functions did not allow a direct representation of some complication
in the plot

The first situation has been easily resolved by allowing the story generation to consider
stories that violate Propp’s premises. Once the roles in the story have been decoupled from
the choice of protagonist, the existing set of character functions allows representation of
different stories simply by shifting the protagonism to characters that do not succeed in the
end. These have always existed as antagonists, and they can now become protagonists of
tragic stories.
The second situation has consequences at two different levels. First, the Proppian set
of character functions did not contemplate complications like fluctuating love relations or
psychological development of characters. The multiplication of the number of possible
schemas for plot arise from the consideration of instances of particular subsequences that
present specific characteristics related to these features not contemplated by Propp. Some of
these complications required a significant overhaul of the expressive power of the underlying
computational system and can only be considered as further work. Yet other complications
would require only a dual process of generalization/instantiation of the character functions in
the existing set to cover the missing features. Propp’s set of character functions was developed
for a very specific set of folk tales and it was not intended to be generalizaded beyond it. The
concept of character function itself, in contrast, was defined as a generic tool for the analysis
of narrative. An extended set of character functions, satisfying Propp’s requirements on
the definition of a character function but covering the range of basic complications outlined
in the present paper would be significant contribution to the field of narrative generation.
The set of character functions developed by Propp has been tested repeatedly as a possible
resource on which to base generic story telling system and has been found wanting [11].
The proposed extension might help to reduce the shortcomings perceived and increase the
expressive potential of system based on a character function representation.
A further extension being contemplated as future work concerns the need for a flexible
mechanism for combining meaningful sequences of character functions into larger narrative
units, which would allow the system to capture more faithfully a larger set of the reviewed
narrative schemas. A grammar-based solution such as the one outlined in [3] is being
considered as a possible solution.

5 Conclusions
A number of existing descriptions of plot has been reviewed, and the resulting analises
have been correlated to distill a basic vocabulary of narrative schemas. These narrative
schemas have been paraphrased in terms of sequences of character functions as described in
Propp’s morphology. This has allowed the extension of an existing story generation system
to generate output stories corresponding to the desired narrative schemas.
Important insights on the expressive power of Propp’s morphology, and some discussion
of its limitations as a generic story generation framework have been outlined. Limitations of
Propp’s morphology have been identified at three different levels. First, the sequencing and
ordering of plot bearing elements/character functions as determined by Propp’s formalism
is too rigid to capture the flexibility of plots beyond Russian folk tales. Second, the set of
abstractions for plot bearing elements/character functions would need to be extend, both
with new elements and with additional annotations to existing ones, for instance regarding
issues like gender of the characters, whether they survive the event, or whether the outcome
is positive or negative for them. Third, an additional level of information concerning affinities
 P. Gervás, C. León and G. Méndez 17

between characters and/or psychological characteristics of the characters may need to be

considered for dealing with Comedy plots as described by Booker or plots of the mind as
described by Tobias.
The work reported in the paper is preliminary and ongoing, and several avenues of future
work have been described. Some of these hold significant potential for improving both the
quality of the resulting stories and the value of the proposed solution as a computational
model of narrative.

Acknowledgements This paper has been partially supported by the project WHIM 611560
funded by the European Commission, Framework Program 7, the ICT theme, and the Future
Emerging Technologies FET program.

References
1 Alexander Nikolayevich Afanasyev. Narodnye russkie skazki A. N. Afanaseva [Folk Russian
tales of A. N. Afanasev], volume 1-3. Moscow, 1855.
2 C. Booker. The Seven Basic Plots: Why We Tell Stories. The Seven Basic Plots: Why We
Tell Stories. Continuum, 2004.
3 Pablo Gervás. Propp’s morphology of the folk tale as a grammar for generation. In Work-
shop on Computational Models of Narrative, a satellite workshop of CogSci 2013: The
35th meeting of the Cognitive Science Society, Universität Hamburg Hamburg, Germany,
08/2013 2013. Schloss Dagstuhl – Leibniz-Zentrum für Informatik GmbH, Dagstuhl Publish-
ing, Saarbrücken/Wadern, Germany, Schloss Dagstuhl – Leibniz-Zentrum für Informatik
GmbH, Dagstuhl Publishing, Saarbrücken/Wadern, Germany.
4 Pablo Gervás. Reviewing Propp’s story generation procedure in the light of computational
creativity. In AISB Symposium on Computational Creativity, AISB-2014, April 1-4 2014,
Goldsmiths, London, UK, 04/2014 2014.
5 S. Imabuchi and T. Ogata. A story generation system based on Propp combined with a
conceptual dictionary. In Natural Language Processing andKnowledge Engineering (NLP-
KE), 2011 7th International Conference on, pages 359–362, Nov 2011.
6 R. Raymond Lang. A declarative model for simple narratives. In Proceedings of the AAAI
Fall Symposium on Narrative Intelligence, pages 134–141. AAAI Press, 1999.
7 G. Polti and L. Ray. The Thirty-six Dramatic Situations. Editor Company, 1916.
8 Vladimir Propp. Morphology of the Folk Tale. Akademija, Leningrad, 1928.
9 D. E. Rumelhart. Notes on a schema for stories. Representation and Understanding: Studies
in Cognitive Science, pages 211–236, 1975.
10 R.B. Tobias. 20 Master Plots: And How to Build Them. F+W Media, 2012.
11 Zach Tomaszewski and Kim Binsted. The limitations of a Propp-based approach to in-
teractive drama. In Intelligent Narrative Technologies, Papers from the 2007 AAAI Fall
Symposium, pages 167–173. AAAI Press, 2007.

CMN 2015

View publication stats