The PModeling Technique: Results Of The CMMIPSPEM Experiment

1
The P-Modeling Technique Results Of The
CMMI-P-SPEM Experiment

• the experiment CMMI-P-SPEM was conducted on
  October 26, 2005, collocated with the First
  Software Engineering Conference in Russia

Vladimir L Pavlov Nikita Boyko
vlpavlov_at_ieee.org mboyko_at_acm.org
2
Agenda

• Historical overview
• October 15 and October 26
• Results of the Experiment

3
The Babel Experiment

• A team of students gets an assignment to create a
  high-level design for a software system
• They are required to use UML and are not allowed
  to use any other language (English, German,
  Russian, Ukrainian, etc), either spoken or
  written
• Students are given 4-6 hours to complete the task
• This assignment was always presented as an
  experiment the students are to discover whether
  or not UML is a real language that can
  successfully serve as a communication tool within
  a team

4
The Babel ExperimentA Training On OOA/OOD

• Developed in 2001by Vladimir L Pavlov
• Since 2001 it has been successfully delivered
  more than ten times to the software developing
  companies and universities
• The results of the experiment were always
  positive students were always able to prove
  that the UML is a real language

5
The Babel ExperimentWas Discussed At

• The First Open Conference IT-Education in
  Russia, Suzdal, May 2003
• http//www.it-education.ru
• 36th ACM Technical Symposium on Computer Science
  Education (SIGCSE 2005), St. Louis, Missouri,
  USA, February 2005
• http//www.ithaca.edu/sigcse2005/
• 18th IEEE Conference on Software Engineering
  Education and Training (CSEET 2005), Ottawa,
  Canada, April 2005
• http//www.site.uottawa.ca/cseet2005/
• These articles and presentations are available
  for download from http//www.vlpavlov.com

6
Once in 2004

• Two teams were working independently on the same
  task
• One team was limited to using only the UML
  language and pantomime in their communication
• The other was allowed to use speech in addition
  to UML
• The first team (which was not allowed to use
  speech) coped with a task more successfully than
  the other team
• Their diagrams were more detailed, more elaborate
  and elegant

7
Not Only Training?

• So, we decided to find out whether the speechless
  modeling sessions can be used in real software
  development projects
• Can they be used not only for training purposes,
  but also to produce real design?
• Is a speechless approach more effective than
  traditional approach?
• The CMMI-P-SPEM experiment was designed to answer
  these questions

8
Agenda

• Historical overview
• October 15 and October 26
• Results of the Experiment

9
The Rehearsal And The Experiment

• The rehearsal was organized to validate and
  correct (if needed) the experiment methodology
• It was organized 10 days before the experiment
• some of the rehearsal participants also attended
  the experiment on October 26
• After the rehearsal, no changes were made to the
  methodology
• The name of the experiment was corrected
• from CMMI-P-SPEM to P-MODELING

10
October 15, 2005Rehearsal

• Was organized in Dnepropetrovsk, Ukraine
• Co-sponsored by StarSoft Labs
• http//www.starsoftlabs.com
• 2 teams, 10 participants

11
October 26, 2005Experiment

• Was organized in Moscow, Russia
• Collocated with the First Software Engineering
  Conference in Russia
• http//www.secr.ru
• 2 teams, 10 participants

12
P-modelingMethodology

• Is based on using
• Silent Modeling Sessions
• a team is not allowed to use speech during the
  modeling session
• Reverse Semantic Traceability
• one team creates a design (translates from
  human language to UML),
• another team validates it (translates it back
  from UML to human language, initial and restored
  texts are compared)

13
Experiment Methodology

• 2 teams work independently on different tasks
• Each team is given a text descriptions of a
  domain, and they have to create a UML model for
  it
• They are not allowed to speak
• They have 3 hours to complete the assignment
• Then the teams exchange the results of their work
  (sets of UML diagrams)
• Based on the received UML models, the teams have
  to restore the original text descriptions
• They are allowed to speak at this phase
• They have 1 hour to complete this task
• Then the teams have a joint workshop where they
  compare results and discuss possible ways to use
  the P-modeling techniques in their real-life work

14
Consolidated Data From Both Rehearsal And
Experiment

• 15 participants
• 12 practitioners
• 2 students
• 1 university instructor
• Experience in Object-Oriented Programming (OOP)
  from 1 to 10 years
• Average 4.5 years
• Experience in Object-Oriented Modeling from 1 to
  5 years
• Average 2.5 years

15
Consolidated Data From Both The Rehearsal And
Experiment

• Participants formed 4 teams
• 4 text descriptions of different domains were
  used
• Russian language was used
• Size from 150 to 300 words (from half of page to
  the whole page)
• These text descriptions were translated to UML
  and then restored back to texts

16
Diagrams

• 3-11 diagrams were created for each model
• Almost all types of UML diagrams were used
• none of the 4 teams created collaboration and
  state-chart diagrams

17
Agenda

• Historical overview
• October 15 and October 26
• Results of the Experiment

18
Results

• In all 4 cases restored texts were very close
  to the original texts
• Almost no data was lost or misinterpreted
• Some new (implementation) data were added
• Most experiment participants said, that the
  silent approach allowed them to create their
  models faster, than it would take using
  traditional approaches
• All experiment participants said that Reverse
  Semantic Traceability is the most powerful design
  validation technique
• 90 of the experiment participants are going to
  use Silent Modeling Sessions in their real-life
  software development projects
• All experiment participants are going to use
  Reverse Semantic Traceability in their real-life
  software development projects

19
Feedback From Participants

• The P-Modeling technique can be used in
  industrial software development for
• Conceptual design
• Increasing effectiveness of current software
  development processes
• Design validation
• Detailed design
• Quick diving into new domains
• Building effective teams

20
The Day Before Yesterday

• The experiment was conducted on October 26 - the
  day before yesterday
• Today is October 28, 2005
• So, we did not have much time to analyze all
  collected data
• A detailed report will be published soon

21
Summary

• Today the P-modeling technique consists of
• Silent Modeling Sessions
• Reverse Semantic Traceability
• The P-modeling technique was successfully used
  before for educational purposes
• The experiment has demonstrated that P-modeling
  can also be used in real-life software
  development projects, and can be more effective
  than traditional modeling approaches

22
Thank You

• RUSSEE
• http//www.russee.com
• StarSoft Labs
• http//www.starsoftlabs.com

23
Thank You

• Ilya Antipov http//www.russee.com
• Alexander Babich http//ppc-kpi.ukrbiz.net
• Dmitry Bednyak http//www.dr.dp.ua
• Dmitry Dakhnovsky http//www.russee.com
• Dmitry Malenko http//www.itgny.com
• Pavel Tsitovich http//www.susu.ac.ru
• Igor Kaftannikov http//www.susu.ac.ru
• Lubov Orlova http//www.russee.com
• Konstantin Philonenko http//www.starsoftlabs.co
  m
• Ivan Poyda http//www.poyda.com

24
This presentation was delivered in Moscow on
October 28, 2005, at the First Software
Engineering Conference in Russia

• You can download it form http//www.secr.ru or
  http//www.vlpavlov.com