Craig Sweet
csweet@comcast.net
|
|
Knowledge / Experience Management:
[1] |
K. Althoff, A. Birk,
S. Hartkopf, W. Muller, M. Nick, D. Surmann and C. Tautz, "Systematic
Population, Utilization and Maintenance of a Repository for Comprehensive
Reuse", in Ruhe, G., and Bomarius, F., editors, Learning
Software Organizations - Methodology and Applications, number 1756
in Lecture Notes in Computer Science, pages 25-50, Springer Verlag,
Heidelberg, Germany, 2000. |
| |
Section 2 provides current state of the art approaches
to reuse. Reuse is not limited to code but can include requirements,
designs, documentation, tool, best practices, technologies, lessons
learned, models (resource, process, product, cost, quality), measurement
plans and data. If items are to be reused, they must be searched
for and similar items retrieved. Case-based reasoning systems extend
the faceted classification, as known in library science, by allowing
the facets to have arbitrary types. Characterization is not trivial
and requires a technique known as domain analysis. This section
is full of good references that should be reviewed further.
Section 3 provides an open tool architecture that
supports reuse and continuous learning. Storage and retrieval are
based on a formalism known as REFSENO. REFSENO provides constructs
to describe concepts, terminal attributes, nonterminal attributes,
similarity functions and integrity rules. The EB schema is specified
using REFSENO's constructs. Figure 2 gives a good architectural
example. It's interesting that there is a separation between the
EB-specific storage system (Case Base stores artifact characterizations)
and the Artifact Specific Storage System (which stores artifacts
in their native formats). So essentially, the EB-specific components
handle characterizations and allow for similarity-based searching
based on those characterizations. Operations specific to
Section 5 provides decompositions of various tasks,
including "learn" and "record". The text clearly
shows that maintaining the EB is an active process that requires
buy-in from everyone. Certainly there will be maintenance tasks
involved. For example, the EB schema may need to be extended. Project
"maui" required such a change. Figure 8 gives an example
of its extended schema. This looks a lot like a relational schema.
My first thought is to think of XSD which describes types and cardinalities.
Not sure if some work can be done on that angle. Also, is there
a general schema that can be applied? That would certainly be helpful
to eliminate the need to extend an existing (and populated) schema.
Also, this schema appears to have a bunch of free-text fields. I
wonder if it would be better to be more limiting.
Section 6 describes many of the roles that may
exist within the EF and their duties and interractions. These include
manager, supporter, engineer, and librarian. This is depicted in
figure 9.
Questions:
What is a semantic relationship
What is context-sensitive retrieval
|
[3] |
V. Basili, "Software
Development: A Paradigm for the Future", COMPSAC '89,
Orlando, Folrida, pp. 471-485, September 1989.
|
| |
This paper begins by exploring some of
the problems software engineers have been struggling with for years.
This involves mechanisms to improve quality and productivity as
well as the dispersement of knowledge. An Improvement Paradigm (IP)
is presented which defines a long term, quality-oriented organizational
life cycle model. The IP has four aspects: Characterizing the environment,
Planning, Analysis, and Learning and Feedback. Each of these four
aspects are further described.
The paper loosely defines the Goal Question Metric
(GQM) paradigm but gives a reference on where to find a set of guidelines.
Most of the remainder of the paper talks about
the experience factory (EF) idea. An EF appears to not be just a
database of solutions but rather a new way to approach intra-organizational
learning. Each project should still be centered around developing
it's products. The EF, which appears to be a mix of people and technology,
is responsible for maintaining the experience base so that its benefit
can be used by many projects. Each project can consult the experience
base as needed throughout the lifecycle but is not necessarily charged
with updating the experience base, that appears to be the job of
the EF.
Again, the EF and experience base do not appear
to be just a database. Ideally this is a set of technologies that
depend on where in the software development lifecycle you are. During
the Characterization IP phase a project manager can query the EF
for information about previous similar projects. This may include
resource and allocation information, personnel experience, available
software and hardware, environmental characteristics and baselines
for schedules. During the planning phase, the EF can help in tailoring
goals via the GQM paradigm. Certainly the output would be different
in this phase. During the execution phase, the EF may provide process
models, methods and techniques and tools. At project conclusion,
the EF may provide benchmarks for comparison.
The paper concluded with some areas of current
(1989) research including automatic code generation and its associated
validation, etc.
|
[4] |
C. Seaman, M. Mendoca,
V. Basili and Y. Kim. "An
Experience Management System for a Software Consulting Organization",
Fraunhofer Center website.
|
| |
This paper begins by giving a great explanation
of a major problem for software developers - applying lessons learned
by others to a new project. The experiences of Q-Labs, Inc. and
their goal to allow each of their consultants to benefit from the
experience of every other Q-Labs consultant, ws presented.
The Visual Query Interface reminds me a lot of the Latent Semantec
Indexing displays that I've seen in literature. I wonder if there
is a correlation?
This paper does not address issues such as database structure or
query structure, protocols, etc. It does provide an interesting
section entitled "Principles Behind the Experience Management
System" which does describe the three levels of an EMS and
its requirements.
|
[5] |
V. Basili, M. Lindvall
and P. Costa, "Implementing
the Experience Factory Concepts as a Set of Experience Bases",
Proceedings of SEKE 2001 Conference, Buenos Aires, Argentina,
June 2001.
|
| |
This paper begins by briefly describing an
the Experience Factory (EF) concept. It is mentioned that this approach
can be used outside of the Software Engineering realm. In fact,
all organizations have a basic need to manage information. This
may be to reduce the impact of employee turnover, employee training,
collecting data about potential new projects and the resources (time,
money, etc.) they require, etc. A list of generic problems facing
all companies is listed on page 1.
The authors state that organizations must change the way they do
business in the future to remain successful. They need to become
less dependent on their employees preventing the loss of knowledge.
Employees may leave or be injured or just plain forget. A company
cannot rely on a few "experts" but rather have a system
in place to collect and disseminate knowledge and experience. Along
this same line, such a system can help new employees become productive
sooner - as a repository of information exists from which they can
draw. Organizations must become 'learning organizations' - where
the sharing of experience, the searching for experience and the
learning from experience becomes a part of the daily life.
It is important for businesses to recognize that mistakes do happen.
The core values of the experience factory make clear that mistakes,
unavoidable as they may be, can be a source of knowledge. Without
a system of collecting that experience and making it available a
company is liable to make those mistakes again. In other words,
those who do not study history are doomed to repeat it.
The paper continues, describing an Experience Management System
as a collection of content, structure, procedures and tools. The
content can be data, information, knowledge or experience, which
is called experience throughout the rest of the paper. The structure
is the way the content is organized. The content and the structure
are often referred to as the Experience Base. Procedures are instructurons
on how to manage the EB on a daily basis, including how to use,
package, delete, integrate and update experience. Finally, tools
support managing the content and the structure.
The paper include a section on steps included
in the methodology. These include characterizing the organization,
defining the user roles, defining a data model (or toxonomy), defining
the architecture, implementing the architecture, deploying the architecture
and maintaining the deployment based on feedback from the field.
The final major section of the paper presents
the results of applying EMS concepts to the Fraunhofer center. The
center itself had to become a learning organization. They created
a set of core values stating upfront what kind of behavior the employees
were expected to adhere to.
Project presentations are discussed. FC-MD
no longer does only post-mortem presentations but regular presentations.
Some initial presentations are one-sided but many are more like
dialogue or brainstorming sessions. These sessions become experience
packages right away. This helps new employees come up to speed on
the projects they are starting on.
To get the EB populated, FC-MD employed answer
gardens, initially populated FAQ's and chat logs.
The Visual Query Interface (VQI) allows searching
through the Experience Base. The VQI maps the experience packages
onto X- and Y- coordinates for easy visualization. The paper did
not address whether the entire EB was displayed or if an "intelligent"
display - perhaps based on LSI - was used. This may be an area of
research. Some mention is made of Hyperwave for document management
and indexing. The authors state that they are working on integrating
hyperwave with EMS.
Mention is also made to a Z drive. It appears
as if all documents are stored on a file share but I may be misunderstanding
the reference. This makes me wonder about EB data storage.
|
[6] |
V. Basili, P. Costa,
M. Lindvall, G. Mendonca, C. Seaman, R. Tesoriero and M.V. Zelkowitz,
"An
Experience Management System for a Software Engineering Research
Organization", Software Engineering Workshop,
NASA/Goddard Software Engineering Laboratory, Greenbelt, MD, November
2001, pp.25-29. |
| |
This paper begins by reiterating the point
that businesses are more reliant on knowledge than ever before.
The problem is that knowledge is owned by employees, not by the
business. That knowledge walks out the door everyday. Most organizations
face this problem but software organizations are particularly hard
hit as tasks are more human and knowledge intensive.
Knowledge Management (KM) is traditionally
viewed as a long-term investment. An investment in either time or
money is made now and benefits are seen later. Employees are also
concerned with the invest now / benefit later approach. Since knowledge
is contained within each employee they are a crucial component.
They will be reluctant to change their processes and start capturing
their knowledge for a payoff that may be received by others. For
these reasons, KM approaches are often seen as risky for managers
and employees alike.
In order to arrive at a quicker benefit from
a KM system, this paper presents the Knowledge Dust to Pearls approach.
It makes use of benefits both from an AnswerGarden and an Experience
Factory. The Answer garden is the knowledge dust and serves short-term
needs. This may be in the form of Question-Answer pairs. The Experience
Factory contains the knowledge pearls and serves the long term needs
of the organizatio. These pearls are often a synthesis of many related
knowledge dust or imini-pearls.
Under this approach, knowledge dust is made
immediately, with minimal modification. In parallel, the knowledge
dust components are analyzed and synthesized into knowledge pearls
by the EF group and placed into the EF.
The implementation of this approach is accomplished
by modifying the traditional EF model. Traditionally every piece
of information (knowledge dust) was analyzed, synthesized, then
packaged in the form of experience packages. Under this new approach,
that process still occurs but a second (quicker) process is also
performed. After the analysis phase, the synthesis phase is skipped
and the results made available (pearls). This allows the the organization
to receive immediate benefit from the knowledge dust, before the
full EF process completes.
Proposed process for qualitative analysis of
experience dust:
- Identify a set of experience packages
- Choose a set of keywords that describe the
topic or issue that needs to be investigated.
- Search the text in the experience packages
for occurrances of the keywords
- Group the coded passages and look for trends
resulting in a "story" or hypothesis, which is a pearl.
- Create an experience package summarizing
the newly created knowledge.
|
[7] |
K. Althoff, B. Decker,
S. Hartkopf, A. Jeditschka, M. Nick and J. Rech, "Experience
Management: The Fraunhofer IESE Experience Factory", Proc.
of the Industrial Conference on Data Mining, Institute for
Computer Vision and Applied Computer Sciences, Leipzig, Germany,
July, 2001.
|
| |
This paper describes some of the basic methods
of Experience Management. This is presented in the context of the
Fraunhofer IESE Experience Factory.
The peper begins with a brief introduction
to Experience Management. This includes information on Case-Based
Reasoning (CBR) and Experience Factories (EF). Clearly EF's use
CBR technology at its core. As the paper mentions, "The underlying
idea of CBR is simple: Do not solve problems from scratch but remember
how you solved a similar problem and appy this knowledge to solve
your current problem". "Since themid nineties CBR is used
both on the organizational EF process level as well as the technical
EB implementation level". An example of an operational EF called
COIN is given in section 3.
Section 4 describes the experience Base Buildup
Method, called DISER. It includes the definition of the six main
steps of an EB development. Section 5 presents the Experience Management
Content Framework (EMCF) as a generic blueprint of an EB.
Section 6 talks about business processes and
capturing lessons learned. It is mentioned that the process modeling
technique of IQ is structured text. The specific structure is not
identified but I thought that an XML approach would
be beneficial. This section also includes interesting information
about capturing and presenting lessons learned.
The peper continues in section 7 with some
mention of maintenance. Processes include a mix of automatic, tool
supported and manually performed maintenance activities.
Section 8 presents an interesting "push"
technique for "subscribing" torelevant EB information.
The paper concludes by investigating data mining in the context
of EM.
|
[8] |
V. Basili, G. Caldiera
and D. Rombach, "The
Experience Factory", Encyclopedia of Software Engineering,
Vol. 1, pp. 469-476, Wiley, 1994. |
| |
This is a very non-technical introduction to
Experience Factories. It begins by outlining the motivation behind
an experience management system. Page 3 provides a nice set of currently
understood software development concepts.
Since EF's use the Quality Improvement Paradigm
(QIP) as its core section 3 provides a basic overview of QIP. Its
six steps (characterize, set goals, choose process, execute, analyze,
package) are examined as well as the control and capitalization
cycles. This section continues by exploring the goal/question/metric
paradigm.
Page 7 provides a brief description of various
life cycle models. Section 4 provides a high level overview of experience
factories. Examples of packaged experience are given in section
5 and include equations, histograms, graphs, lessons learned, models
and algorithms.
The paper concludes by giving examples and
implications of EF's.
|
[9] |
K. Althoff, U. Becker-Kornstaedt,
B. Decker, A. Klotz, E. Leopold, J. Rech and A. Voss, "Enhancing
Experience Management and Process Learning with Moderated Discourses:
The indiGo Approach", Proc. of European Conference
on Artificial Intelligence (ECAI '02) Workshop on Knowledge Management
and Organizational Memory, 2002. |
| |
This paper describes an approach to creating
and sustaining living process models. This is done via discourses
- deliverative, reasoned communication focused and intended to cluminate
in decision making. These moderated discourses are made available
and text mining approaches are run. Examples are made of process
models that do not quite fit and updates or tangents are made by
way of these discourses
This paper, hopefully in draft, is poorly written
and quite confusing. It fails to clearly explain how moderated discourses
are incorporated into the EF flow.
|
[10] |
V. Basili, "The
Experience Factory and its Relationship to Other Improvement Paradigms",
4th European Software Engineering Conference (ESEC), Garmish-Partenkirchen,
Germany, September 1993. |
| |
Unable to find
|
[11] |
V. Basili and G.
Caldiera, "Improve Software Quality by Reusing Knowledge and
Experience", Sloan Management Review, Vol. 37,
pp. 55-64, 1995.
|
| |
Unable to find
|
[15] |
M. Lindval, I. Rus
and S. Sinha, "Software Systems Support for Knowledge Management,"
Journal of Knowledge Management, Vol. 7(5), pp. 137-150,
2003. |
| |
Unable to find. Should tie together SE and
KM
|
[16] |
K. Althoff, M. Nick
and C. Tautz, "Improving
Organizational Memories Through User Feedback", Proc.
of the Workshop on Learning Softwre Organizations at SEKE '99,
Springer, Kaiserslautern, Germany, June 1999, pp. 27-44. |
| |
This paper begins by making the argument that user
feedback (in an organizational memory) is essential if that OM is
to be beneficial in the future. It presents a goal oriented method
called OMI (Organizational Memory Improvement) which improves an
OM incrementally from the user's point of view. At each OMI step,
protocol cases are recorded to pinpoint improvement potential for
increasing the perceived usefulness. If an improvement potential
is identified the user is asked for specific improvement suggestions.
This allows the OM to adapt to the needs of the users even if the
environment changes.
Section 3 provides interesting information on perceived
usefulness of query results. It shows that there are many factors
which (including user urgency, organizational climate, tool familarity,
etc.) that influence the perceived usefulness. The several areas
that can affect the perceived usefulness are clearly explained.
Section 4 shows the "ideal" sequential
usage model of an OMMS. It is argued that this ideal situation never
occurs. This may have to do with the way people query for information.
People often start queries but then seek additional information
or refine their query based on the information retrieved.
Section 5 shows a usage model that incorporates
user feedback. This may be quite a bit of work for the user though.
The user is asked to rank the responses in order and evaluate in
that order. Throughout this process ths user is asked to provide
information. If a candidate is rejected, they should provide information
on why it was useless. Based on this feedback the OM maintenance
functions can look to fill in any "holes".
The paper continues by demonstrating the utilization
of protocol cases using GQM plans as an example. This is a topic
that I have not yet fully explored.
|
[17] |
R. Weber, D. Aha,
N. Sandhu, H. Munoz-Avila, "A
Textual Case-Based Reasoning Framework for Knowledge Management
Applications ", Knowledge Management by Case-Based
Reasoning: Experience Management as Reuse of Knowledge (GWCBR 2001).. |
| |
This paper introduces a textual case-based reasoning
system (TCBR) framework for KM systems that manipulates organizational
knowledge embedded in artifacts. The TCBR approach acquires knowledge
from human users and from text documents using template-based information
extraction methods, a subset of natural language, and a domain ontology.
Their approach is similar to [18] but is not specific
to the software engineering discipline
I have only glanced through this paper but it does
show other research in combining KM and CBR.
|
[18] |
K-D Althoff, F. Bomarius,
and C. Tautz, "Using
Case-Based Reasoning Technology to Build Learning Software Organizations",
in Proceedings of the 1st Workshop on Building, Maintaining,
and Using Organizational Memories (OM-98), Brighton, UK, August
1998, Volume CEUR-WS/Vol-14.. |
| |
notes
|
[20] |
L. Briand, C. Differding
and H. Rombach, "Practical
Guidelines for Measurement-Based Process Improvement",
Software Process, 2(4):253-280, December 1996.. |
| |
This paper provides practical guidelines for planning,
implementing and using goal-oriented software measurement for process
improvement. It seeks to provide more guidance to people performing
measurement programs using the GQM paradigm
It provides a motivation for goal-oriented measurement,
processes to follow, definitions of goals, construction of a GQM
plan and subsequent analysis.
I have only skimmed this article but it may require
further reading if my research takes me in that direction.
|
|
 |
|
