Content uploaded by Gonzalo Méndez
All content in this area was uploaded by Gonzalo Méndez on Jun 27, 2018
Content may be subject to copyright.
INES: A reconstruction of the Charade storytelling system
using the Afanasyev Framework
on and Pablo Gerv´
as and Gonzalo M´
Facultad de Inform´
Instituto de Tecnolog´
ıa del Conocimiento
Universidad Complutense de Madrid
The present paper introduces INES (Interactive Narrative
Emotional Storyteller), an instance of the Afanasyev story
generation framework that rebuilds Charade, an agent-based
storytelling system. The construction of INES pursues a dou-
ble goal: to develop a more complete version of Charade, by
including a plot generation stage; and to show the capability
of Afanasyev as scaffolding for building united systems from
sources of diverse kind. From a broad view, the resulting
architecture is a microservice-oriented ecosystem in which
every signiﬁcant stage of the story generation process is im-
plemented by a microservice that can be easily replaced by
another, as long as the new microservice keeps the interface
contract established by the Afanasyev model.
Automatic story generation is a part of a wider research area
in Artiﬁcial Intelligence named Computational Creativity
(CC), which is the pursuit of creative behaviour in machines
A story generator algorithm (SGA) refers to a computa-
tional procedure resulting in an artefact that can be consid-
ered a story (Gerv´
as 2012). The term story generation sys-
tem can be considered as a synonym of storytelling systems,
that is, a computational system designed to tell stories.
The operation of the story generation systems requires
large amounts of knowledge. These systems are faced with
a signiﬁcant challenge of acquiring knowledge resources in
the particular representation formats that they use. They
meet an inherent difﬁculty when using formal languages
in the detachment between the formulation of the needs in
the real world and its representation in a formal construc-
tion. A possible solution can be the use of a Controlled
Natural Language (CNL) for knowledge interchange (Con-
on et al. 2016). This is precisely the approach intro-
duced by the Afanasyev framework (Concepci´
endez 2017a; 2017b). Afanasyev is a collaborative
architectural model for automatic story generation which re-
lates to a service-oriented architecture. It introduces an ag-
nostic story representation model (Concepci´
endez 2017b) that intends to ease the collaborative inter-
change of knowledge between different systems.
INES (Interactive Narrative Emotional Storyteller) is a re-
construction of Charade (M´
as, and Le´
based on the Afanasyev Framework. The original Charade
system is a simulation-oriented agent-based story genera-
tion system. Charade was focused on generating stories
about the evolution of the relationships between characters
by running an unrestricted low-level simulation. The de-
velopment of INES introduces a new stage in the Charade
generation model, that is the plot generation. This stage pro-
vides the system with a more structured way of building the
stories. Also, the development of INES allows for testing
the suitability of the Afanasyev architectural structure and
its knowledge representation model in a real-world context.
The ﬁrst story generation systems date back to the 1970s.
The Automatic Novel Writer (Klein 1973) is considered to
be the ﬁrst storytelling system. It generated murder stories
in a weekend party setting by means of generation gram-
mars. TALE-SPIN (Meehan 1977) was another of the ear-
lier story generators. It generated stories about the inhabi-
tants of a forest. TALE-SPIN was a planning solver system
that wrote up a story narrating the steps performed by the
characters for achieving their goals. Author (Dehn 1981)
was the ﬁrst story generator to include the authors goals as a
part of the story generation process. To this end, it intended
to emulate the mind of a writer. From a technical point of
view, Author also was a planner but, unlike TALE-SPIN, it
used the planning to fulﬁll authorial goals instead of char-
acter goals. Universe (Lebowitz 1984) generated the scripts
of a TV soap opera episodes in which a large cast of charac-
ters played out multiple, simultaneous, overlapping stories
that never ended. In contrast with Author, Universe gave a
special importance to the creation of characters, as it con-
sidered they were the driving force for generating stories.
Brutus (Bringsjord and Ferrucci 1999) was a system that
generated short stories using betrayal as leitmotiv. The main
contribution of Brutus was its rich logical model for repre-
senting betrayal. This feature, along with its grammar-based
generation component and its literary beautiﬁer allowed it
to generate quite complex stories. The Virtual Storyteller
(Faas 2002; Swartjes 2006) is a Multi-Agent System that can
generate stories by simulating a virtual world in which char-
acters modeled by agents pursue their goals. In this way, the
story emerges from the events in the virtual world. Fabulist
(Riedl and Young 2010) is a complete architecture for auto-
matic story generation and presentation. Fabulist combines
an author-centric approach together with a representation of
Despite there is not much speciﬁc literature on the subject,
there are some noticeable efforts concerning the reconstruc-
tion of an existing story generation that have been carried
out. Minstrel (Turner 1993) was a story generation system
that told stories about King Arthur and his Knights of the
Round Table. Each story was focused on a moral, which
also provided the seed for developing the story. Minstrel
was develop in Lisp (Berkeley and Bobrow 1966) and used
an extension of a Lisp library called Rhapsody (Malkewitz
and Iurgel 2006) for representing the knowledge required by
the generation process.
Skald (Tearse et al. 2014) is a publicly-released rational
reconstruction of Minstrel for analysing original Turner’s
work in search of new implications for future research.
Skald is written in Scala, a functional programming lan-
guage that runs over the Java Virtual Machine. It is based
on a previous project named Minstrel remixed (Tearse et
al. 2012), that tried to develop a collection of improvements
over the original Minstrel. The original components of Min-
strel, as described in Turner’s dissertation (Turner 1993), are
the starting point for the Skald design. The work developed
in Skald can be considered not only a collection of enhance-
ments over Minstrel but a globally different picture of the
original Minstrel, as well as a new system that sets the stage
for future research in story generation.
One of the Skald key ﬁndings is an improvement over
Minstrel’s limitations: the story library, story templates, and
the recall system must be tailored to one another for the orig-
inal system to function. Tearse (2012; 2014) shows that this
can be mitigated through a number of techniques, such as
adding differential costs to transformations to remove the
least-successful author-level actions.
Although Skald contains a good number of enhancements
over Minstrel remixed, these are all aimed to expand its ca-
pabilities in a few areas: transparency in story generation,
exploration and measurement of the subtle workings of in-
dividual modules, improved stability, and better story output
in terms of speed, size, and coherence.
Skald keeps the original Minstrel speciﬁcations, in the
sense that it simulates the actions of a human when pro-
ducing stories. Skald puts its novelty at a lower level, us-
ing different levels of modules for simulating different prob-
lems. For example, it uses low level simulations of problem
solving processes, while authorial goals are simulated using
modules in a higher level.
Materials and methods
as, and Le ´
on 2014; 2016) models
the relationship between two characters using their mutual
afﬁnities, and applies it for generating stories.
This system is an agent-based architecture developed us-
ing JADE (Java Agent Development Framework) (Bellifem-
ine, Poggi, and Rimassa 1999). It consists of two types of
agents: a Director Agent, that sets up the execution environ-
ment and creates the characters; and the Character Agents,
one for each character of the story, whose interactions gen-
erate the story.
The main objective of the system were implementing an
afﬁnity model as decoupled as possible from the story do-
main, and testing it independently from other factors such as
the environment in which the action takes place or the per-
sonality traits and emotional state of the characters. Due to
this independence, it can be easily used to generate different
kinds of stories.
The generator is based on a simulation of the characters’
interaction. During the simulation, the characters perform
actions that result in a variation of their afﬁnity levels. Ac-
cording to the afﬁnity level, the characters can be a couple,
friends, indifferent, and enemies. Generation is indepen-
dent of the domain; although, since it focuses on afﬁnities, it
works best in domains where this afﬁnity makes sense. The
simulation is not directed, so that it can not be considered
to constitute a plot or a story by itself. The input includes
a complete parametrization of possible actions, categorized
according the type of relationship allowed for the characters,
the simulated characters, and their relationships measured in
terms of afﬁnity. The output consists of a list of actions pro-
posed by characters, and the response of their counterparts,
that can accepted or reject the proposals, with the variation
of afﬁnity between the characters involved. Despite no text
is generated, it would be easy to use a template for generat-
ing a textual description.
as, and M´
2017b) is a framework speciﬁcally designed for building
service-based automatic story generation systems. From an
architectural point of view, it is basically a collection of mi-
croservices orchestrated by a high-level service. Each ser-
vice exposes their capabilities as REST-based APIs (Field-
ing 2000) and it understands and generates JSON messages.
Due to the fact that the inner logic of any microservice can
come from a different storytelling system, its interface must
be adapted to match the required contract so the microser-
vice can operate under the conditions speciﬁed by the frame-
work. This is the reason why Afanasyev includes the deﬁ-
nition of the common REST interfaces provided by the ser-
vices and leaves to every particular system the details of the
The main microservices in Afanasyev, depicted in Figure
1, are the following:
•Story Director, the microservice that orchestrates the
•Plot Generator, the microservice that generates a high-
•Episode Generator, the microservice that ﬁlls the scenes
that composed the plot.
•Filter Manager, the microservice that manages a set of
ﬁlters that will be applied to the story each time it changes
(due to the activity of the Episode Generator).
Figure 1: Architecture of Afanasyev.
•Draft Reﬂector, the microservice that analyses the story
for deciding whether it is completed or not.
•Discourse generation services (Discourse Planner, Sen-
tence Planner and Linguistic Realization), which turn the
abstract story model into a human-readable text in Natural
In order to allow the combined operation, the microser-
vices of the framework require a common representation
model for stories. The Afanasyev representation model
as, and M´
endez 2017b) focuses on the
knowledge that is directly related to the story, instead of
that related to the generation process, which would be hard
to export between different systems. The model has been
designed as a hierarchical structure, in which the root con-
cept is the story. Most of the leafs of this tree-like structure
are asserts representing a piece of knowledge. These asserts
are expressed by means of sentences in a Controlled Natu-
ral Language (CNL) (Schwitter 2010). In Afanasyev, every
story is composed by a plot and a space. The plot repre-
sents the sequence of events —actions and happenings, that
constitutes the skeleton of the story. The space encompasses
the whole universe in which the story takes place, includ-
ing the existents —characters, living beings and objects that
take part in the story, and the setting —the set of locations
mentioned in the story.
Persistence in Afanasyev is mainly composed by two
stores: the Draft Repository and the Knowledge Base. The
Draft Repository is a database that stores the ongoing drafts.
The current implementation of this component is based on
a NoSQL database (Han et al. 2011), namely MongoDB
(2017). The knowledge base has the task of preserving all
the knowledge related to concepts, relationships between
concepts, rules, etc. It is a knowledge base generated from
the contributions of the involved story generation systems.
This model of knowledge syndication allows to increase the
Figure 2: Architecture of INES.
shared set of concepts each time a new system joins the
ecosystem. Hence, every contributor performs an initial load
expressing its rules.
INES is the translation of the Charade storytelling system to
the Afanasyev architectural framework. The purpose of this
work is two-fold: to validate the capability of the Afanasyev
model for supporting different story generation models and
to prepare the integration of Charade in a wider service-
based collaboration ecosystem.
The main adaptation work has focused on a central aspect
of the original Charade behaviour: the directed simulation.
In effect, Charade originally produced outputs that were the
result of an unrestricted simulation. In the case of INES,
there is a preexisting plot to which the output of every sim-
ulation must be adapted. This means that, for each scene,
there is a speciﬁcation based on precondition / postcondition
that implies that not every possible result of the simulation
The architecture of INES, as adapted from Afanasyev, is
depicted in Figure 2. It is a combination of Afanasyev ready-
made services, along with the speciﬁc Charade adapted ser-
vices and a set of newly created services, required by the
•Story Director, provided by Afanasyev
•Plot Generator, required by Afanasyev and newly devel-
oped for INES
Figure 3: Operation of INES / Afanasyev.
•Episode Generator, created from the Charade system
•Emotional tension ﬁlter, created from the Charade system
•Draft Reﬂector, provided by Afanasyev
•Text generator, required by Afanasyev and newly created
The Story Director
The architecture of Afanasyev is an ecosystem of microser-
vices. The Story Director manages the joint operation of the
whole ecosystem, as depicted in Figure 3. It orchestrates
the execution of the different story generation stages by re-
questing the APIs of the different services. This processing
proceeds iteratively, generating drafts that will be reﬁned in
each pass, until the established criteria for story complete-
ness are met.
The ﬁrst step consists in generating the basic structure of
the plot. It is performed by the Plot Generator, that estab-
lishes the sequence of episodes that make up the plot. Each
episode is interwoven with the others by means of its pre and
post-conditions. These are collections of statements relating
to the setting and the existents of the story.
The Plot Generator: Audrey
The Plot Generator, named “Audrey” —after Audrey Hep-
burn who played the lead role in “Charade”, has been de-
veloped speciﬁcally for INES and is a template-based plot
generator which produces outlines from a subset of the
cinematographic basic plots compiled by Ball´
erez 2007). Its basic procedure can be considered as sim-
ilar to systems like Gester (Pemberton 1989) and Teatrix
(Machado, Paiva, and Brna 2001). The basic idea behind
Audrey is building a story plot containing the main scenes
that will be developed by the Charade-based Episode Gen-
erator. The plot building procedure starts by selecting one
of the persisted templates, and then it gives it substance by
instantiating the generic elements of the template.
An example of one of these templates is “The destructive
outsider”. This story is essentially composed by the follow-
•The initial state: a peaceful community.
•The arrival of the outsider.
•The outsider acts against the members of the community,
performing destructive actions, without being uncovered.
•The true evil nature of the outsider is revealed.
•The heroes rise from the community and ﬁght against the
•The outsider is purged. The community becomes peaceful
In order to developed a consistent detail for every episode,
INES requires a knowledge base that contains the main con-
cepts presented in the plot. In this case, the plot mentions a
“community”, an “outsider”, some “destructive actions” per-
formed by the outsider, a group of “heroes” that rise against
the outsider, and certain “purging actions” that the heroes
perform. All these concepts are related to each other and
can be represented by means of a graph. So, the required
knowledge for instantiating the example is partially depicted
in Figure 4.
So, the relationships between the concepts can be consid-
ered as asserts such as: “When the community is a family,
then the outsider can be a new partner, an unknown relative
and a new lover. When the outsider is a new partner, then
the arrival can be a marriage”.
The translation of these relationships to a physical
database ﬁts better with a graph-oriented database. In this
particular case, the knowledge base has been implemented
using Neo4j (Vukotic et al. 2014). So there are nodes with
labels such as “Community”, “Outsider”, “Arrival” and “Ac-
tion” representing the main plot concepts. The relationships
between the instances of the concepts are represented by
means of the graph edges (i.e. database connections between
The next step concerning this knowledge is to apply it for
determining the actions that the characters can perform dur-
ing the simulation in order to keep the story consistency. For
example, the knowledge base can label as “hostile actions”
the acts of “insult” and “kill”. If the outsider has harmed
the community by sowing discord, it would be unjustiﬁably
Figure 4: Partial view of the concepts relationships in the KB.
excessive that the heroes reacted by killing him. In this case,
the context of the plot should provide the proper actions to
The Episode Generator
This microservice is based on the original Charade core. It
generates a complete simulation of characters interaction ac-
cording to the restrictions that are provided as input param-
eters. As mentioned in the previous section, the types of
actions that can be consistently performed by the characters
are limited by the context of the story. So, the episode gen-
erator receives not only the ongoing story, but also the pre-
conditions and postconditions that the resulting simulation
must match. This approach introduces a shift in the prior
behaviour of the Charade’s engine, which originally drove
an unrestricted simulation.
Charade was designed for obtaining the list of possible
actions from its conﬁguration, during system startup. It dis-
tinguished between three types of actions (love, friendship
and enmity). In order to ease the adaptation of Charade as
a microservice in the Afanasyev ecosystem, the set of possi-
ble actions are passed to it as part of the request parameters.
For selecting the most suitable actions, the Story Director
queries the knowledge base and retrieves the context-related
actions that better ﬁt the storyline. For example, continu-
ing the previous example, if the plot is related to a fam-
ily and the outsider has stolen something, the actions car-
ried out in response to this offence could be “insult”, “re-
port the burglary”, “demand the restitution” and “demand to
leave”. These options would be retrieved and passed in the
request as the proper actions that could be performed by the
“heroes” of the story. Then, the episode generator would se-
lect some of them during the simulation and complete the
detail of the episode.
Table 1 shows a sample story that can be generated by
applying this model.
The Emotional tension ﬁlter
The current version of the Emotional tension ﬁlter works in
a very simple way. It is a ﬁlter which is invoked after every
episode simulation —performed by the Episode Generator,
and it determines if the generated actions ﬁt certain drama
parameters. To meet this purpose, the Emotional tension
ﬁlter considers the semantic information associated to the
actions in the knowledge base to adjust the strength of the
drama in the story. For example, considering again the story
of “The destructive outsider” plot, an action such as “to slap”
the outsider is much more dramatic than “to demand him to
leave”. By establishing the threshold for the tension, this
service helps the Story Director to select the most dramatic
continuation of the plot. So, this ﬁlter removes a subset of
the generated episodes, and makes the Story Director to call
John invites William to dinner
John invites Mary to dinner
William helps John to cook
Mary gives a present to John
The arrival of the
John makes a welcome party for
David (the outsider)
David gives a present to John
William helps David to move
Mary helps David to move
David steals a valuable object
from John’s house
David tells Mary that William is
Conﬂict Mary believes David
Mary insults William
William gets angry with Mary
The outsider revealed William discovers David steal-
ing in John’s house
William tells John that David is
John tells Mary that David is the
The rise of the heroes John insults David
John demands David to leave
David leaves the town
Conclusion Mary says sorry to William
William gives thanks to John
Mary gives thanks to John
Table 1: A sample story based on “The destructive outsider”
again the Episode Generator until the whole plot has been
The Draft Reﬂector
The Draft Reﬂector of INES is the original basic Afanasyev-
provided Draft Reﬂector. This microservice simply checks
if all the episodes have been developed according to the plot
restrictions. In this case, the choice is based on the need
of keeping the draft analysis stage as simple as possible.
The interest of the INES model is related to the ability of
Afanasyev to provide a suitable architecture for building a
system like Charade by means of its building blocks.
The Text Generator: Lampert
The other INES-speciﬁc service developed is the Text Gen-
erator, named “Lampert” —after Audrey Hepburn charac-
ter’s surname in “Charade”. Lampert is microservice that
translates the plot, represented as a data structure, into a text
in Natural Language. Its core is based in the SimpleNLG
Java library (Gatt and Reiter 2009).
The text generation is the last stage in the story generation
process. Lampert has been designed simply for conveying
the story represented in the Afanasyev common representa-
tion. Its purpose is not so much being a literary beautiﬁer
but providing a human-readable summary of the story.
Afanasyev is not focused towards the ad hoc integration of
speciﬁc pre-existing systems, but rather to provide a gen-
eral service-oriented framework that allows the construction
of different storytelling systems by assembling components
from various systems (or from only one, in the simplest
case). For this reason, the adaptation of Charade has re-
quired a number of transformations. Firstly, Charade has
been designed as a lightweight agent-based architecture. Its
essential logic has been preserved in INES, but the system
operation has been restructured. The operation of every mi-
croservice in INES is completely independent from the oth-
ers. Adapting the simulation ﬂow has involved the devel-
opment of a couple of new components that did not ex-
ist in Charade: the Plot Generator and the Text Generator.
These two microservices could be easily replaced by other
microservices based on different approaches that the current
ones, as this is the essence of the Afanasyev framework.
The generation model of the Plot Generator is quite sim-
ple, but also very convenient for ﬁlling the existing gaps in
the original model. It can be easily extended by providing
more plot structure templates. Also, the richer the knowl-
edge base is, the more interesting the generated stories are.
As it has been shown, the role of the knowledge base is es-
sential in this model for achieving coherent and believable
stories. The same basic plot template can be instantiated in
a wide spectrum of stories. As new instances are added to
the database, the variability will increase accordingly.
Another relevant addition to the original Charade be-
haviour is the Emotional Tension Filter. It allows the sys-
tem to generate stories with a greater drama, or not, depend-
ing on the ﬁltering values. This service can be enhanced,
or even replaced by a much more complex one, in order to
help to create stories according to certain narrative tension
curves. This conﬁguration will entail a more global way of
operation, considering not only the particular tension of an
episode, but the evolution of the whole narrative arc.
Despite its simplicity, the Text Generator service provides
a useful output. It has been deliberately designed for provid-
ing a summary in Natural Language rather than an elaborate
literary text. Naturally, it can also be replaced by a much
more complex surface realizer which provides a more pol-
ished literary work. A future candidate could be the TAP
SurReal Surface Realizer (Herv´
as and Gerv´
Conclusions and future work
The Afanasyev framework, despite having been originally
conceived as an architectural model for building collabora-
tive storytelling architectures, should not be seen solely as
a tool for system integration. The purpose of developing
INES was to prove that the Afanasyev framework can also
be used for rebuilding any system as a microservice-based
model. In this particular case, the adaptation of a purely
agent-based simulation-oriented story generation system to
a microservice-based pre-existing framework was particu-
larly challenging. The resulting system can be considered
an evolved version of the original Charade system, with a
more structured approach to story generation.
Another interesting derivative of the work carried out dur-
ing the design and development of INES is the knowledge
base itself. It has been addressed using a representation
model based on a graph-oriented database. This has allowed
for simplifying the representation, as well as to use a gen-
eral industry-oriented development stack, instead of a stack
speciﬁcally oriented to Artiﬁcial Intelligence. The fact that
the database can also be consulted by means of a REST in-
terface provides an additional decoupling mechanism that
will allow to evolve it independently, and even its replace-
ment, without affecting the operation of the rest of the mi-
In the present version of INES, for every draft processed
in every iteration, several continuations can be generated and
added to the population of drafts to process during the next
iteration. On the generated population, a reﬂection process
is applied by means of the Draft Reﬂector microservice, and
the drafts that it considers already ﬁnished are marked as
stories. This process continues until all drafts are marked
as ﬁnished or a limit of iterations is reached (to guarantee
completion). In the face of future work, the development of
a service that helps to decide what is the most appropriate
level of detail in each of the scenes is still pending. This
aspect can be provided in a ﬁrst instance by a human —
applying a co-creation model—, but it would be perfectly
evolved to introduce a component for automating this task.
This paper has been partially funded by the projects IDi-
LyCo: Digital Inclusion, Language and Communication,
Grant. No. TIN2015-66655-R (MINECO/FEDER) and
InVITAR-IA: Infraestructuras para la Visibilizaci´
on y Transferencia de Aplicaciones y Resultados de In-
teligencia Artiﬁcial, UCM Grant. No. FEI-EU-17-23.
o, J., and P ´
erez, X. 2007. La semilla inmortal: los
argumentos universales en el cine. Ed. Anagrama.
Bellifemine, F.; Poggi, A.; and Rimassa, G. 1999. Jade–a
ﬁpa-compliant agent framework. In Proceedings of PAAM,
volume 99, 33. London.
Berkeley, E. C., and Bobrow, D. G. 1966. The programming
language LISP: Its operation and applications. MIT Press.
Bringsjord, S., and Ferrucci, D. 1999. Artiﬁcial intelligence
and literary creativity: Inside the mind of brutus, a story-
telling machine. Psychology Press.
on, E.; Gerv´
as, P.; M´
endez, G.; and Le´
on, C. 2016.
Using cnl for knowledge elicitation and exchange across
story generation systems. In International Workshop on
Controlled Natural Language, 81–91. Springer.
on, E.; Gerv´
as, P.; and M´
endez, G. 2017a. An
api-based approach to co-creation in automatic storytelling.
In 6th International Workshop on Computational Creativity,
Concept Invention, and General Intelligence. C3GI 2017.
on, E.; Gerv´
as, P.; and M´
endez, G. 2017b. A com-
mon model for representing stories in automatic storytelling.
In 6th International Workshop on Computational Creativity,
Concept Invention, and General Intelligence. C3GI 2017.
Dehn, N. 1981. Story generation after tale-spin. In IJCAI,
volume 81, 16–18.
Faas, S. 2002. Virtual storyteller: an approach to computa-
tional storytelling. Unpublished masters thesis, University
of Twente, Department of Electrical Engineering, Mathe-
matics and Computer Science.
Fielding, R. T. 2000. Architectural styles and the design of
network-based software architectures. Ph.D. Dissertation,
University of California, Irvine.
Gatt, A., and Reiter, E. 2009. Simplenlg: A realisation
engine for practical applications. In Proceedings of the 12th
European Workshop on Natural Language Generation, 90–
93. Association for Computational Linguistics.
as, P. 2012. Story generator algorithms. In The Living
Handbook of Narratology. Hamburg University Press.
Han, J.; Haihong, E.; Le, G.; and Du, J. 2011. Survey on
nosql database. In Pervasive computing and applications
(ICPCA), 2011 6th international conference on, 363–366.
as, R., and Gerv´
as, P. 2009. Evolutionary and case-
based approaches to reg: Nil-ucm-evotap, nil-ucm-valuescbr
and nil-ucm-evocbr. In Proceedings of the 12th European
Workshop on Natural Language Generation, 187–188. As-
sociation for Computational Linguistics.
Klein, S. 1973. Automatic novel writer: A status report.
Papers in text analysis and text description.
Lebowitz, M. 1984. Creating characters in a story-telling
universe. Poetics 13(3):171–194.
Machado, I.; Paiva, A.; and Brna, P. 2001. Real characters in
virtual stories. In International Conference on Virtual Sto-
rytelling, 127–134. Springer.
Malkewitz, S. G. R., and Iurgel, I. 2006. Technologies for
interactive digital storytelling and entertainment. In TIDSE.
Meehan, J. R. 1977. Tale-spin, an interactive program that
writes stories. In In Proceedings of the Fifth International
Joint Conference on Artiﬁcial Intelligence, 91–98.
endez, G.; Gerv´
as, P.; and Le ´
on, C. 2014. A model of
character afﬁnity for agent-based story generation. In 9th
International Conference on Knowledge, Information and
Creativity Support Systems, Limassol, Cyprus, volume 11,
endez, G.; Gerv´
as, P.; and Le´
on, C. 2016. On the use of
character afﬁnities for story plot generation. In Knowledge,
Information and Creativity Support Systems. Springer. 211–
2017. Mongodb ofﬁcial site. https://www.mongodb.
com/. [Online; accessed 29-December-2017].
Pemberton, L. 1989. A modular approach to story genera-
tion. In Proceedings of the fourth conference on European
chapter of the Association for Computational Linguistics,
217–224. Association for Computational Linguistics.
Riedl, M. O., and Young, R. M. 2010. Narrative planning:
balancing plot and character. Journal of Artiﬁcial Intelli-
gence Research 39(1):217–268.
Schwitter, R. 2010. Controlled natural languages for knowl-
edge representation. In Proceedings of the 23rd Interna-
tional Conference on Computational Linguistics: Posters,
COLING ’10, 1113–1121. Stroudsburg, PA, USA: Associa-
tion for Computational Linguistics.
Swartjes, I. 2006. The plot thickens: bringing structure and
meaning into automated story generation.
Tearse, B. R.; Mawhorter, P. A.; Mateas, M.; and Wardrip-
Fruin, N. 2012. Lessons learned from a rational reconstruc-
tion of minstrel. In AAAI.
Tearse, B.; Mawhorter, P.; Mateas, M.; and Wardrip-Fruin,
N. 2014. Skald: minstrel reconstructed. IEEE Transactions
on Computational Intelligence and AI in Games 6(2):156–
Turner, S. R. 1993. Minstrel: A Computer Model of Cre-
ativity and Storytelling. Ph.D. Dissertation, University of
California at Los Angeles, Los Angeles, CA, USA. UMI
Order no. GAX93-19933.
Veale, T. 2013. Creativity as a web service: A vision of hu-
man and computer creativity in the web era. In AAAI Spring
Symposium: Creativity and (Early) Cognitive Development.
Vukotic, A.; Watt, N.; Abedrabbo, T.; Fox, D.; and Partner,
J. 2014. Neo4j in action. Manning Publications Co.