COLLABORATIVE BUSINESS PROCESS ELICITATION
THROUGH GROUP STORYTELLING
João Carlos de A. R. Gonçalves, Flávia Santoro and Fernanda Baião
NP2TEC – Research and Practice Group in Information Technology, Department of Applied Informatics
Federal University of the State of Rio de Janeiro (UNIRIO), Rio de Janeiro, Brazil
Keywords: Business Process Modelling, Knowledge Elicitation, Computer-Supported Collaborative Work.
Abstract: Business Process Modelling remains a costly and complex task for most organizations. One of the main
difficulties lies on the process elicitation phase, where the process analyst attempts to extract information
from the process’ participants and other resources involved. This paper describes a case study in which a
previously proposed Story Mining method was applied. The Story Mining method and its supporting tool,
ProcessTeller, makes use of collaborative storytelling and natural language processing techniques for a
semi-automatic extraction of BPMN-compliant business process elements from text.
1 INTRODUCTION
Business Process Modeling is a very time-
consuming and expensive task. Nevertheless, it is
vital for most organizations, since it enables people
to explicit and to share their knowledge about how
business tasks are actually performed in day-to-day
activities. The main factor that brings so much
difficulty to this area is the knowledge issue.
Knowledge can be defined as “the combination
of data and information to which expert opinion,
skills and experience are added” (Davenport and
Prusak, 1998). There are two main types of
knowledge: explicit knowledge, externalized and
able to be used, and tacit knowledge, very difficult
to register or externalize (Nonaka and Takeuchi,
1995).
One of the main problems with business
modeling resides on the fact that the source of
knowledge is usually the activity performers of the
process to be modeled. Even if documentation or
other sources of information are available, generally
they are incomplete or outdated. Although many
techniques were developed and used for extracting
the information from the performers and participants
of the organization’s tasks, process modeling (and
specially process elicitation) remains complex.
One of the most widely-used techniques for
process and knowledge elicitation is the interview.
Based on a series of questions chosen by the analyst,
who focuses on the relevant knowledge for the
modeling and elicitation of a given process, the
analyst interviews people involved with the tasks
that are executed during the process. Later, the
analyst transcripts the answers and analyzes them, in
order to extract the process knowledge.
Nevertheless, the interview technique presents
some drawbacks. First, both the selection of the
questions and the interviewees can be biased.
Second, further interpretation of the interview
transcript by the analyst will be inevitably restricted
to his viewpoint about the entire process. Alvarez
(2002), during a research involving the analysis of
requirements elicitation based on interviews,
evidences the possibility of incomplete knowledge
from the user due to the analyst’s bias.
Based on previous research (Leal et al., 2004;
Freitas et al., 2003), we argue that collaborative
technique based on storytelling is more effective in
collecting knowledge about work process. Thus, we
propose a Group Storytelling approach, supported by
the Story Mining method (Gonçalves et al., 2009)
and a groupware tool, named ProcessTeller. A case
study applying this tool in a real environment was
conducted and its results are presented in this paper.
This paper is structured as follows: Section 2
briefly describes the Story Mining method, Section 3
presents ProcessTeller tool, Section 4 describes the
case study and Section 5 concludes the paper.
295
Carlos de A. R. Gonçalves J., Santoro F. and Baião F. (2010).
COLLABORATIVE BUSINESS PROCESS ELICITATION THROUGH GROUP STORYTELLING .
In Proceedings of the 12th International Conference on Enterprise Information Systems - Information Systems Analysis and Specification, pages
295-300
DOI: 10.5220/0002910002950300
Copyright
c
 SciTePress

2 A METHOD FOR
COLLABORATIVE PROCESS
ELICITATION
Process Discovery techniques and methods can be
classified in two distinct groups: (i) the first group
emphasize the “human factor” of business process,
focusing on the people that participate on the
activities regarding a given process; (ii) the second
group focus on the automated and non-human
factors of a process, including mining process from
information system logs and automatic techniques
for knowledge acquisition. Both groups have
different advantages and disadvantages.
We proposed the Story Mining method
(Gonçalves et al., 2009) which intends to bridge the
gap between these two groups, making it able to
access the richness of knowledge present at the
process’ participant, while tapping on the swiftness
of automatic proposals. Leal et al. (2004) states that
a story is a natural way to make the knowledge
explicit. They proposed the Group Storytelling
technique, which is based on collective narrative,
and the construction of stories by groups of people
for knowledge elicitation in organizations. Based on
free-form narrative, knowledge about different
subjects can be found on a story. We argue that its
application for process elicitation enhances its
results. However, the large amount of collected
information that is not related to the process may
become a problem.
As a solution to this new problem, the use of
Text Mining and Natural Language Processing
(NLP) techniques was proposed, in order to provide
support for the analyst at the interpretation of the
story’s content and the structuring of process
elements in a process model. The proposed Story
Mining method is divided into three main phases.
The first phase is the storytelling part, where the
tellers are chosen and the collaborative process
happens. After the story is collectively created, the
second phase applies NLP and Text Mining
techniques on the story text elements. The
techniques involved at this stage are out of scope of
this paper and will not be described. The third and
final phase automatically builds a formal
representation of the process from the story. Using
the elements extracted by the method’s second phase
as a guide, the analyst can model the process using
the desired notation and tools, and discusses it with
the participants for further improvements and,
hopefully, reach a consensus about how the process
should be depicted.
In order to support the Story Mining method, the
ProcessTeller tool was created, and is described in
the next section.
3 THE PROCESS TELLER TOOL
Group Storytelling is a technique based on
collaborative free-form narrative, aiming at
collective constructing and sharing knowledge
among the storytellers and other participants.
This section describes ProcessTeller, a
supporting tool for the Story Mining method
previously proposed in (Gonçalves et al., 2009).
ProcessTeller is a web application, implemented
as an extension of the TellStory tool (Leal et al.,
2004), for process-oriented Group Storytelling that
allows users to collaboratively create stories. Stories
are composed of a chain of events (textual excerpts
of the story). It allows alternative flows of events in
the same story, additional story elements (such as
documents linked to events) and group management
features (such as polls).
The new functionalities of ProcessTeller are
described below.
3.1 Story Groups
The original tool allowed the user to list all stories,
regardless of theme or other criteria. As process
elicitation is focused on a “theme” (the process
itself), the tool was modified in order to show a
directory structure of stories classified by Story
Groups.
3.2 Document Linking to Story Events
Tellstory allows documents related to the story being
told to be stored and linked to it. However, in order
to benefit from Text Mining and NLP techniques as
much as possible, ProcessTeller enables its users to
provide specific information about every excerpt of
a process description, by linking documents to each
of the story events. Users are now able to upload
relevant documents and determine which events are
related to each document.
3.3 Event-level Character Detection
Characters are important elements of a story, having
active roles during the description of a narrative.
They come closer to the performer of a task in a
business process (which may be represented as lanes
in BPMN (BPMN, 2010) notation).
ICEIS 2010 - 12th International Conference on Enterprise Information Systems
296

In order to improve the quality of the story being
told, while avoiding direct interference with the
storytelling collaborative process, at the time of a
new event’s insertion, if its text does not contain
known characters, ProcessTeller detects it and
suggests the user to register a new character or
modify the event’s content.
3.4 Starting Event
A story that describes a business process is not just a
narrative, it usually has a starting event that enables
it to happen. Nevertheless, free-form stories told by
users may not present events in chronological order,
or may not be initiated by its first event. Therefore,
ProcessTeller enables its users to arbitrarily establish
the starting event of a story. A Start Event is defined
as “where a particular process will start” (BPMN,
2010). Thus, a new attribute “Starting Event” was
added to each story, so as to enable the
ProcessTeller to know.
4 CASE STUDY AT DIA/UNIRIO
4.1 Description
A Case Study using ProcessTeller was conducted at
the Department of Applied Informatics (DIA) of the
Federal University of the State of Rio de Janeiro
(UNIRIO). The case study aimed to evaluate the
viability of the Story Mining method (Gonçalves et
al., 2009) and its supporting tool, ProcessTeller, as
well as to extract evidences and new insights about
the knowledge issues involving the collaborative
storytelling process.
The chosen process to be modelled was Course
Enrollment, since it is an extremely common process
in educational institutions and known by a broad
range of “storyteller candidates”, including students,
professors and university staff.
Although there were different processes at the
institution, due to the presence of bachelor and
master’s degree courses at the same department, a
decision was made to not divide the two processes in
two different narratives but, instead, to allow
participants of both contexts to tell their viewpoints
and experiences in a single story, in order to assess
the richness of knowledge and the differences
present in the narrative.
Invitations to participate on the case study were
sent through e-mail to 18 people, including
undergraduate students, university staff, graduated
students (at the Master’s Degree level) and
professors. Additionally, an open invitation was sent
to the mailing lists of all students and professors.
During a full month, the users told their
experiences in Course Enrollment using
ProcessTeller. They were also able to read and
comment each other´s contributions, include
additional elements such as characters, upload new
documents and relate them to the story events.
At the first week of the case study, the first user,
a bachelor course student, created 8 new events,
describing his entire view of the process. The
analysis of the case study results pointed out that
those 8 initial events functioned as a “skeleton view”
of the basic process components, as other
participants commented his events and added new
ones, complementing the story as a whole.
As the case study progressed, additional 18 story
events were added to the main story flow, and 51
new comments were made by the users. As the
narrative grew larger and richer, some users have
chosen to contribute only with their comments on
previously created events, leaving their opinions and
insights on other participants’ contributions. After
the end of the second week of the study, the number
of event creations slowed down and the rate of
adding new comments on existing events increased.
At the end of the method’s first phase, the main
flow of the story told by the participants and the
auxiliary information registered (characters,
documents, among others) were used for the second
phase, the application of text mining and natural
language processing algorithms.
This stage generated the proto-model in two
notations, BPMN (BPMN, 2010) and XPDL (XPDL,
2010), therefore two output files were generated.
BPMN was chosen since it is an OMG standard,
while XPDL was used due to the fact that it is the
visualization format adopted by the most popular
Business Process Modelling tools. Also, an
additional XML file was generated by ProcessTeller,
in which each process element (Actor, Activity and
Parameter) is described in more detail.
The files generated by ProcessTeller should not
be taken as the final version of the process model
but, instead, as an intermediary version to be
validated and improved. The graphical visualization
of the generated model enables the analysis to easily
assess the knowledge present at the story and to
modify it to achieve the final representation.
Meanwhile, these files may also be used by story
participants to visualize the exposed process. This
variety of ways for the analyst to assess the captured
knowledge and achieve the final model composed
COLLABORATIVE BUSINESS PROCESS ELICITATION THROUGH GROUP STORYTELLING
297

the third and final phase of the method, where a final
formal representation of the process is built.
At the end of the case study, a questionnaire was
applied to the participants, where they reported their
opinions and perceptions about the experience of
using the ProcessTeller tool and how close they
found the final story was to the Course Enrollment
process.
4.2 Analysis of Results
The results achieved by the method can be divided
in two groups: The extracted activities by the
method (Table 1 and Table 2), and the analysis of
the Questionnaire answers from the case study
participants (Table 3).
The final process model resulted from the case
study (which we will call PM1) was compared to
another version of the process model, which was
manually created using interviews (PM2 model), in
order to evaluate the precision of the automated
knowledge extraction. The results of this comparison
are in Table 1, depicting the coincident and non-
coincident activities between the two models.
Table 2 groups the extracted activities and
classifies them in three groups, based on thieir
content: (i) General activities (common activities of
a Course Enrollment process); (ii) Master’s Degree
activities (specific activities belonging to Master’s
degree course enrollment); and (iii) Bachelor’s
Degree activities (specific activities belonging to
Bachelor’s degree course enrollment) and “Special”
Students (activities belonging to the course
enrollment of visiting students at UNIRIO).
Table 1: Process models comparison.
Statistics PM1 PM2
Total # of activities 21 51
Total # of coincident activities 8 21
Total # of non-coincident activities 13 30
Table 2: Extracted activities by group.
Group # of Activities
General 35
Bachelor’s degree 4
Master’s degree 6
“Special” Student 6
Total of # Activities 51
The main difference between the Story Mining
activities and the pre-existent model activities are
the perceptions of the people involved. For instance,
activities that are related to the usage of information
systems or that are mainly administrative were
depicted on the traditional model, while activities
that were clearly visualized by the participants were
present at the Story Mining’s activities.
A number of suggestions can be raised from this
observation about the activities. Although characters
are already presented on a story, special elements of
the process, such as systems and business rules, can
be described in different ways, stimulating the tellers
to describe activities involving them. Table 2 also
shows that there is a tendency for an increasing
capture of activities common to a “generic” Course
Enrollment process, composed by activities like
“Student selects a course” and “Professor offers
course”. However, the activities related to
alternative flows (for example, activities specific to
each course type) are also captured as well, even if a
smaller number of participants may have been in
contact with them. The answers from the
questionnaire are summarized in Table 3. A selected
number of questions were evaluated, regarding the
participants’ opinions about several characteristics
of the study and the tool itself.
Table 3: Questionnaire results.
Question Agree Indifferent Disagree
The tool allowed the
expression of your
viewpoint regarding
course enrollment?
10 1 1
Telling a story in a
collaborative way
allowed the easy
expressions of your
viewpoint about course
enrollment?
9 2 1
The event document
attachment functionality
was useful to complete
your viewpoint of the
event?
7 5 0
Automatic detection of
character-less events
stimulated the register of
new characters?
6 4 2
Polling was useful to
solve incoherence and
conflicts?
2 10 0
Event operation "Switch
Places" was useful for a
better flow of story
events?
9 2 1
Event operation "Join
Events" was useful for a
better flow of story
events?
9 3 0
Event operation "Break
Event" was useful for a
better flow of story
events?
9 3 0
ICEIS 2010 - 12th International Conference on Enterprise Information Systems
298

Table 3 shows a concise view of the case study.
The first questions were very important, since they
are related to the validity of the method for process
elicitation and the ProcessTeller tool. The majority
of users confirmed its usefulness, while a small
group disagreed. The main causes of disagreement
were: the blend between master’s degree and
bachelor’s degree course enrollment events on the
same story and the inability, at a case study level, for
each user to tell a separate story.
Three questions had less than 75% approval (i.e.
9 replies on “Agree”) from the users. The first two
regards character detection and the attachment of
documents to events. They may be explained due to
the reduced number of new events as long as the
story developed and the process depicted achieved
its completion, as some users weren’t able to
experience the character detection feature of
ProcessTeller, explaining the high number of
“Indifferent replies”.
The last question is about the poll feature and has
the highest number of “Indifferent” replies as well as
zero “Disagree”. This fact points out that polls were
not used during the case study and the users
arguably preferred to use comments to discuss issues
related to the events.
Also, three open questions were present,
regarding their viewpoint about the process depicted
and the users’ contribution to the story, the final
story generated, the tool itself and other general
comments about the elicitation task as a whole.
Specific trends were observed from the users, based
on their answers to these questions:
Many users could not assert that the story
reflected the complete or correct Course Enrollment
process, due to the fact that they had a limited
viewpoint of it. But, on the other hand, they were
confident that their contributions were correct and
that their experiences with the process were reflected
on it.
Another group of users expressed their surprise
with details about the process that they were not
aware of, because they had contact with just a small
part of its activities.
An abundance of knowledge was also noticed by
the participants, as some of them stated that “there is
a mix of master degree and bachelor degree events”
at the story. The majority of them agreed that they
expressed their limited knowledge, but were
satisfied by the final story, as being the reunion of
these smaller contributions in a collaborative way.
The fact, highlighted at the case study’s
description, of a single user registering many events,
triggered an increase of contributions (new events,
comments, as well as new characters and
documents). It reinforces the value of the
collaborative element of Group Storytelling, as
people probably would not recall so many process
elements, without the aid of the first user’s narrative
elements.
Finally, the increasing usage of event comments
was unexpected, specially in cases where the
comment feature was used as an “ad-hoc forum” for
discussions on a specific event’s contents.
5 RELATED WORK
Indulska et al. (2009) carried out a study in order to
identify a research agenda for process modeling.
They include the following items: the value of
process modeling and expectations of stakeholder
groups involved in process modeling. One of their
conclusions was that group design is an
advantageous approach.
Ryan and Heavey (2006) argue four
requirements for a collaborative approach in process
modeling: have a low modeling cognitive load and
therefore be capable of being used by non-
specialists; present modeling information at a high
semantic level so that personnel can rationalize with
it; have good visualization capabilities; and, support
project teamwork. Our approach is aligned with all
of them, although that graphic visualization will
only occur at the end of the interaction when model
is generated.
Freitas et al. (2003) propose a cooperative
graphic editor (CEPE - Cooperative Editor for
Processes Elicitation) that supports the building of
the knowledge about the current process and intends
to support the reporting of associated problems.
6 CONCLUSIONS
The case study conducted and described in this work
suggests that the Group Storytelling technique may
be successfully applied for process elicitation. The
participants’ responses about the extracted process
support this observation, since most of them agreed
that the story reflected the desired process.
The main advantages of the proposed technique,
compared to the traditional approach, is the
lessening of analyst’s bias as well as the free
expression of its own knowledge by the participants,
regardless of their level of involvement with the
process.
COLLABORATIVE BUSINESS PROCESS ELICITATION THROUGH GROUP STORYTELLING
299

Additional features could improve the tool and
the elicitation method, as deeper detailing on
characters, the use of more complex relationships
between events and the linking of extracted activities
and story events, making it able for the analyst to
cross-reference the method output with the original
part of the story.
For the Text Mining and NLP techniques phase
of the method, the increased usage of text comments
on event, surpassing the number of events of the
story brings up the need for a future consideration on
the comments’ text use as input for these algorithms,
broadening the range of its application from the
story events to all textual elements available.
ACKNOWLEDGEMENTS
The authors would like to thank Petrobras for
supporting this project. Flávia Maria Santoro is
partially supported by CNPq (Brazil) under the grant
305404/2008-3 and João Carlos de A. R. Gonçalves
is supported by CAPES.
REFERENCES
Alvarez R., 2002. Discourse Analysis of Requirements
and Knowledge Elicitation Interviews. In Proc. 35th
Hawaii International Conference on System Sciences.
BPMN – Business Process Modeling Notation,
http://www.bpmn.org/Documents. Last accessed on
January 2010.
Davenport, T. and Prusak L., 1998. Working Knowledge,
arvard Business School Press.
Gonçalves, J. C. A. R., Santoro, F. M., Baião, F. A, 2009
Business Process Mining from Group Stories. In:
Proceedings of 13th International Conference on
Computer Supported Cooperative Work in Design,
161-166.
Leal, R. P., Borges, M. R. B., Santoro, F. M., 2004.
Applying Group Storytelling in Knowledge
Management. In IX International Workshop on
Groupware (CRIWG), Lecture Notes in Computer
Science 3198 , 34-41
Freitas, R., Borges, M. R. B., Santoro, F. M., Pino, J. A.,
2003. Groupware Support for Cooperative Process
Elicitation. Lecture Notes in Computer Science, Paris,
v. 2806, 232-246.
Indulska, M., Recker, J., Rosemann, M., Green, P, 2009.
Business Process Modeling: Current Issues and Future
Challenges. Advanced Information Systems
Engineering, 501-514.
Nonaka, I. and Takeuchi H., 1995. The Knowledge-
Creating Company: How Japanese Companies Create
the Dynamics of Innovation, Oxford University Press.
Ryan, J., Heavey, C., 2006. Process modeling for
simulation. Computers in Industry, 57 (5), 437-450.
XPDL – XML Process Definition Language.
http://www.wfmc.org/xpdl.html, last accessed on
January 2010.
ICEIS 2010 - 12th International Conference on Enterprise Information Systems
300