Question-Answer system for object-oriented analysis and design

P.I. Sosnin & E.P. Sosnina

Ulianovsk State Technical University, Ulianovsk, Russia

ABSTRACT: This paper describes the results of studies and implementations which can add basic Rational
Unified Process (RUP) flows by a special workflow to support "Interaction with experience”. Such workflow
is developed with orientation on RUP on the basis of a special NetWIQA system that supports question-
answer reasoning of a design team. NetWIQA allows designers to create the drafts of the basic RUP samples
for their subsequent coding and use in RUP environment or without reference to RUP; set of samples based
on the protocol of QA reasoning, its analysis and necessary transformations. NetWIQA is the client-server
QA processor with the special command system.

1 INTRODUCTION a special flow to support "interaction with experi-

ence”. Such workflow is developed with orientation
The Rational Unified Process (Kroll & Kruchten, on RUP (but irrespectively of RUP) on the basis of a
2003) and relative systems of object-oriented analy- special NetWIQA system that supports question-
sis and design (OOAD) characterize the modern ap- answer reasoning (QAR) of a design team in a cor-
proach to Computer-Aided Design. Such means give porate network.
designers an interactive access to useful patterns and
Construction Informatics Digital Library http://itc.scix.net

sample actions, and it is used to manage the design

activities and to borrow the necessary samples with 2 RELATED WORK
the purpose of their rational implementation in the
process of designing and its results. The domain of our study includes the similar works
The RUP gives us the access to normative models with experience, knowledge and its models, e.g. Ra-
of design experience and knowledge prepared to be tional Unified Process and similar techniques of
used in a corporate network. The models of experi- automated design; Question-Answer techniques and
ence in RUP are the richest source of questions, e.g. systems; presentation and processing of experience
questions checking conformity of the project to the in Case-Based Reasoning approach to computer-
appropriate standards; questions, which provide per- aided design.
sonal and collaborative understanding, rational deci- RUP and other OOAD toolkits effectively solve
sion-making, and etc. Such questions are invariant to the task of processing of patterns and templates
the contents of the project, but are very important in (Galic at al., 2000), i.e. every pattern being the cer-
designing. tain model of experience presents one of the typical
Unfortunately, RUP has no effective means for project decisions to be applied in other projects. At
current work with questions and answers presented the same time RUP structure has no means to sup-
as human reasoning, i.e. reasoning of designers. At port the work with personal knowledge and experi-
the same time reasoning is the basic way of interac- ence of the every member of a team and collabora-
tion of people with the design experience and its tive knowledge. The concept and term “Experience”
models. is not practically used in RUP documents and pa-
Those OOAD systems, in which interaction with pers.
the experience is realized obviously, e.g. in BORE Last years we have observed a growing interest to
system (Henninger, 2003), use methods and means Question-Answer (QA) techniques and systems.
of access only to typical decisions and are based on Many important scientific and practical results were
Case-Based Reasoning (CBR) approach. obtained in several domains, e.g. in retrieval of rele-
Below we submit the results of our studies and vant information in distributed, multimedia, multi-
implementations which can add basic RUP flows by lingual and multi-agency data sets; in automated
 learning in traditional and distance education; in The basic function of Knowledge, being the main
automated design and decision-making in the differ- part in the structure of Experience (Fig. 1), is its or-
ent areas. But despite more than 40 year studies “we dering, which increases efficiency in accessing to
are just beginning to explore question-answering as the components of experience, e.g. rational skills
a research area” (Hirschman & Gaizauskas, 2001). and automatic skills. We consider that the studies in
The main directions of QA research are named in the AI domain should be based on Intelligence and Ex-
Roadmap Paper (Burger et al., 2001) and focused perience as the key concepts.
on: question taxonomies, its understanding, ambi- 2. We claim that question-answering, being a
guities, and reformulations; context and data sources natural function in intelligence, is an effective way
of QA; real time and interactive QA; advanced rea- to access experience for its application and devel-
soning and user profiling for QA. opment, e.g. for automated design.
At the same time the directory of IBM papers The question is understood by the author as a spe-
(www.redbooks.ibm.com) mentions the concept QA, cific natural/artificial phenomenon active when it is
or Question-Answering, only 4 times and doesn’t necessary to apply experience. The concept of
mention the QA work with human experience at all. “question” is understood and defined as a mismatch
Question-Answering systems are also imple- between the potential experience, necessary to im-
mented in the work with knowledge and experience plement a project, and the real experience of a de-
in Case-Based Reasoning techniques (Yang at al., sign team.
2003; Han at al., 2003; Branting at al., 2004). We The question controls the access to experience
consider these approach to be one of the main solu- and its development through the answer and the
tions, which provide the effective modeling and process of answering, and such management is the
processing of knowledge and experience in Artificial essence of a question. We consider that if Case-
Intelligence domain. Based Reasoning is used in forms of Question-
At last, we would like to mention the system Answer Reasoning while designing, it will provide
BORE (Building an Organizational Repository of naturalness in interaction of designers with experi-
Experience) as one of the most related toolkits ence and its models. Positive functional potential of
(Henninger, 2003) to our study, but BORE applica- QAR is much higher, than functional potential of
tion of QA is limited to the set of pre-conditions for CBR.
working with the necessary rules of technique. 3. We believe that studies in Computer-Aided
Design and the solutions are the great source of
tasks and ideas to develop Artificial Intelligence.
3 INITIAL STATEMENTS During design, experience and intelligence are used
Construction Informatics Digital Library http://itc.scix.net

naturally.
Our purpose to develop automated means of interac- Here presented studies and techniques are based
tion with experience is caused by a number of rea- on the following hypothesis:
sons, e.g.: QA interaction with experience consists of natu-
1. We are sure that modeling of experience is ral and artificial components being a complex unit.
much more important than modeling of knowledge Such interaction may be much more effective if:
for future development of Artificial Intelligence. 1 Question-Answer Reasoning of the designers is
To prove it we should realize that human intelli- step-by-step registered in a special database of
gence appeared while evolution, and was an answer the project.
to a question "What can add the genetic forms of ex- 2 The results of QAR registering (QA protocol) are
perience?" While its evolution, Intelligence got the investigated, transformed to useful models, and
functions of self-organization, self-development, used to support the next steps of interaction with
transfer and use of out-genetic components of ex- experience in designing.
perience.
Person It has a number of useful interpretations such as:
1 QA protocol registers any design process as a re-
Experience search experiment, so representing “the primary
Knowledge
measuring information” about design process and
about the used control facilities.
2 Contents of the protocol reflect a real reasoning,
Rational skills Automatic which can be investigated to increase the knowl-
skills edge about a “phenomenon of reasoning and
questions”.
3 Each of the registered questions and answers ad-
mits its interpretation as an event essential for
reasoning and design process, that allows to con-
Figure 1 Experience and knowledge
 sider the protocol as “a network of events” or- design team. NetWIQA has a rich set of operations
dered in time. to interact with all models of experience.
4 Protocol is a data structure (QA structure) with its
practically useful set of operations. Such interpre-
tation of QA structure corresponds to the experi- 5 QA PROCESSOR
ence of Computer Science in the area of data
structuring for adequate presentation of a QA First of all processor NetWIQA is a system for sup-
protocol. porting of the collaborative design process. This sys-
tem gives the members of design team an opportu-
nity of QA recording of the step-by-step designing
4 COMPARISON WITH RUP to the uniform QA protocol.
NetWIQA is the client-server processor with the
To indicate the place and role of our NetWIQA sys- special command system with the following func-
tem in OOAD framework, we present its possible tions:
association with RUP. − to make QA protocols of QAR;
Scheme on Figure 2 generally shows that Net- − to represent the protocols or their parts as the ap-
WIQA allows designers to create: plied models (dynamically visualized and con-
− drafts of the basic RUP samples (for example, all trolled graphs; models of prototypes for sample
kinds of UML diagrams: Use-Case, Classes, Col- actions; the arguments, motivation and purposes);
laboration, Activity and so on) for their subse- − to transform the protocols to network structures
quent coding and use in RUP environment or (QA event nets, QA Pert nets and QA Petri nets);
other related OOAD means; − to analyze texts of questions and answers (the
− drafts of the basic RUP samples for their applica- analysis is aimed at decoding and representing of
tion in the current project without reference to semantics of reasoning with first-order predicate
RUP (using only the possibilities of NetWIQA); logic).
− a set of samples based on the protocol of QA rea-
soning, its analysis and necessary transformations Representations: Transformations:
(such samples are presented below). :
Visualization QA protocols QA event
schemes nets
Patterns, templates and
Other units of RUP Experience Other needed Motivation and
Experience purposes QA Pert
Construction Informatics Digital Library http://itc.scix.net

schemes diagram
R
Design team Precedent
schemes QA Petri
nets
A Argumentation
schemes
Net WIQA
P Analysis of text

Means of registering Questions {Qi} Ontology of a

and editing project

Symple predicates Predicate

formulae
Artifacts of RUP Artifacts of
Net WIQA
Semantic graph schemes
Project

Figure 3 A set of QAR artifacts

Figure 2 RUP and NetWIQA QA processor has two set of the commands, e.g.
instruction dialogue through the visual list of alter-
The system is based both on RUP models of ex- native commands; and plug-in mechanism that pro-
perience and personal experience of the members of
 vides an access to default commands and new acces- 6 QA TECHNOLOGY
sible functions.
These operations of QA processor are enough for QA command system supports the workflow "Inter-
development both the basic UML and RUP samples, actions with Experience" as the special set of appli-
and QAR samples (Fig. 3). In more detail, our proc- cations, each of which perform the certain task of
essor is developed according to the component ap- QA technique. At present we have developed about
proach as the 3-layer architecture on the basis of 40 utilities that are accessible to designers through
Borland MIDAS technology. WEB-environment of a client workstation. The
It consists of layers: structure and the access to the service utilities are
1 QA database and Experience database (Libraries realized according to the interactive RUP-model.
of QA templates and other samples of experi- Each service task is submitted as:
ence), − its QA pattern that are copied to QA database of
2 business-logic kernel (server), the project and applied at the next case of the ser-
3 presentation layer ("thin client”). vice task;
The business-logic layer provides direct data ac- − interactive form to support its convenient applica-
cess for the clients, and performs all data processing. tion.
Server component of QA processor consists of an To demonstrate the functions of these applica-
application server, system magazine, synchronizer tions let us name a few of them, e.g. "To solve a
and etc. connected through TCP/IP sockets with task", " To answer a question", "To construct the an-
Borland Socket Manager utility. swer", "To prove the answer", "To discuss the an-
The third presentation layer of the system archi- swer", " To fill-in the pattern", "To set the require-
tecture involves a client library, a library of syn- ment", " To generate the specification", " To
chronizing flows, a library of configuration, a client determine the front of work", "To plan", "To control
workstation, and the configuration utility to adjust visualization of QAR fragment", "To generate the
users, working groups and projects. The client work- educational unit", "To control knowledge". The
station uses plug-ins to connect necessary functional given utilities show that some applications of QA
modules to expand the set of QA applications, and in processor support also operative training for the
such a way additional applications to work with de- members of a design team, e.g. to learn the certain
sign experience and knowledge are built-in QA en- design process.
vironment. QA technology is applied in any project as coor-
QA processor has the distributed and shared dinated performance of the set of utilities and tasks
structure. The separate components can function on appointed to the members of a design team. At any
Construction Informatics Digital Library http://itc.scix.net

different computers in the local business network. time each designer usually works with the dynamic
QA processor should ensure automated perform- set of the service and project tasks.
ance of the following main tasks: The common work begins with the utility "Con-
− effective information and control support in ceptual design", which calls the following task "To
Computer-Aided Conceptual Design of the auto- open the project" as the initial condition of QA data-
mated systems; base of the project. The initial condition of a task-
− current unified documenting of processes and tree is submitted in Figure 4.
products, open for intellectual interaction; QA templates of documents
Z.0
− development of data and knowledge bases of Z.0.1. < Feasibility Study Report>
complex projects and their interactive use while Z.0.2. <Conception>
the conceptual design; Z.0.3. < Statement of Work for Design >
− controlled forming and accumulation of employ- Z.0.4. <Technical report of the project>
ees experience as QA databases of a design or- Z.1 Z.1.0 Service tasks for:
ganization;
− control of automated designing, e.g. due to the ra- < Defining the text of the main task>
tional choice of the direction of processes while <Prototype of the main task>
<Motivation and purposes schemes>
design decision-making; Other service tasks
− controlled interaction between designers in col- Z.1.1 The first iteration
laborative work in the structure of a corporate de-
sign network; Z1.1.0
− communicative document circulation in the cor- Z1.1.1
porate network;
− controlled training and learning during designing Z1.1.2
and operation of complex systems.
Z.1.2 The second iteration
 Figure 4 Task structure of the project Questions and answers get symbolic codes and
The standard QA patterns of further documenting indexes appropriate to their types and place in the
are loaded at the initial state of QA database. So, e.g. project. Questions have two subtypes, i.e. Z-task and
the State Russian Standards require that any Q-quiry. Q-quiries require the answers to be got
CAD/CAM/CAE system includes the standard from experience database. If any Q has not got the
documents, i.e. "Conception" (analogue of "Vision" answer, it is changed to Z-subtype.
in RUP), "Statement of Work for Design" and "Fea- The answer class is wider and includes subtypes
sibility Study Report". Each pattern of documenting "A-answer", "I-idea", "H-hypothesis", "S-
represents the system of normative questions which specification", "R-requirement", "M-motive", "O-
should be copied to QA database and answered dur- purpose", "E-decision", "P-procedure" and "F-
ing design process. formula".
The system service utilities provide "interaction The QA structure of any task is a dynamic data
with experience" for the definite project tasks, which structure composed during the decision-making and
are formulated and solved during design. reasoning. There are 3 types of questioning activity:
At any moment, design and service tasks make a − automatic reference to experience "automatic
uniform graph, where each task is interpreted as a skill"
“question -task". − rational reference to experience, e.g. with a ques-
Each task is coded by a symbol "Z" and index tion of Q-subtype;
appropriate to a place of this task in the project (in- − rational reference to experience, e.g. with a ques-
dex "0" specifies that a task has utility functions). tion of Z-subtype;
Indexes allow designers to determine not only inner
project tasks from utilities, but also the tasks of dif- The model of QA process is extended to the event
ferent iterations of the project. QA networks and other network models of coherent
Each task in a tree can be in one of the following sets of questions and answers. It requires to solve the
conditions - "decided/solved", "unsolvable", "post- tasks of detection of questions, their identification,
poned" and "executed". Any task has the set of at- adequate coding and modeling, and effective appli-
tributes with references to its templates in QA data- cation. For these purposes QA processor has the
base. means of text analysis with automated translation to
Each project work is supported by QAR while PROLOG-like and semantic net languages.
decision of the task. NetWIQA step-by-step proto-
cols all questions and answers and their relations
(Fig. 5) in QA database of the project. Here we see 7 QA PROCESS
Construction Informatics Digital Library http://itc.scix.net

only QA relations, while the database also includes

other relations (for example such as temporary, per- At definite time ti the reasoning QAR(t) goes to
sonification and cause relations). QAR(ti)-state, which has its causal potential that
gives the possibility to move the reasoning forward
Q to the next state QAR(ti+1). In this aspect the “his-
Q1 A1 tory” of previously made work, represented in
Q11 A11 QAR(t)-codes, influences the next rational step of
reasoning. Next steps both for reasoning and for de-
Q12 A12 sign can be defined by means of question-answer
analysis of QAR(t)-codes.
Q1m A1m General statement of each project task should be
Q2
defined before Question-Answer working with this
A2
task. Special definition of the task (as its general
Q21 A21
statement) uses a special pattern to present a task as
Q22 A22 3 structured text blocks.
The first reflects the main purpose of a system
Q2n A2n under design, which is specified by its potential us-
ers. Here we begin the work with the basic Use-Case
diagram for the task in UML language.
Qp Ap The second defines the main techniques to per-
Ap1
Qp1 form Use-Case diagram for the task. It provides con-
Ap2 struction of the basic diagram of business -objects of
Qp2 UML.
Qpr Apr The third defines technology of implementation
of a system under design. Information of this block
Figure 5 Question-answer structure of the task is applied in conceptual design as context informa-
tion.
 Analysis of a text T0 of the general statement of a 8 CONCLUSION
task and its translation to PROLOG-like language
are used for extraction of questions to begin and In this paper we present the additional step in devel-
continue QAR. opment of the application toolkit and technique for
More detail it is based on step by step registering OOAD. This step includes the special workflow “In-
of questions and answers in accordance with follow- teraction with Experience” applicable to the RUP
ing technique: scheme of OOAD. Such workflow is based on spe-
1 The set of questions {Qi} is taken from the text cially organized QAR, its protocol as primary design
Ò0 and coded by adequate texts T (Qi. information and processing for effective conceptual
2 Actions of item 1 are executed for each text T design in collaborative environments.
(Qi, therefore the set of questions {Qij} and their Methods and means offered, investigated and re-
codes {T(Qij} is formed. Actions of item 2 are alized by the authors open some new possibilities to
used to control the correctness of question codes control CAD processes such as: effective monitoring
and for the choice of those questions {Qk} which of a QA process; analysis of opportunities of parallel
will be used for the next step of detailization from coordination of work in a design team (with the pur-
the set Q = {Qi}∪{Qij}. Other questions are re- pose of distribution of work between the executives
corded for their application in the subsequent in typical design decisions); demonstration (at a suit-
steps. able speed) of the development of events at the cer-
3 Set of answers {Ak} and their codes {T(Ak} is tain time interval of automated design; demonstra-
formed and registered in QA database. tion of the current condition of automated design at
4 Each text T(Ak is processed as the text Ò0. The the certain moment; training to the typical decisions
cycle 1-4 is repeated until the project comes to (CAD samples) and development of design skills;
the end. personification of events for subsequent definition
The work with questions, answers, QAR and of authorship and contribution of the members of a
RUP artifacts are controlled with the help of the ap- design team.
propriate system utilities.
The number of especially important actions (op-
REFERENCES
erations, command, plug-in mechanism, application
of QA processor includes: Branting L., Lester J., Mott M. 2004. Dialogue Management
1 For questions: detection of obvious questions (on for Conversational Case-Based Reasoning. Proc. of the
their indicators, predication (through translation Seventh European Conference on Case-Based Reasoning
on Prolog-like language, identification (on pat- (ECCBR 2004), Madrid, Spain: 65-71.
Construction Informatics Digital Library http://itc.scix.net

terns, concrete definition (for types, assignment Burger J. et al 2001. Issues, Tasks and Program Structures to
of meanings to attributes (as to the phenomenon Roadmap Research in Question & Answering (QA). NIST.
Galic M., Macisaac B., Popescu D. Using a Single Business
of event type, argumentation. Pattern with the Rational Unified Process.
2 For answers: creation, assignment of a type, http://www.redbooks.ibm.com/, REDP-3877-00.
change of a type, registration of a condition, edit- Han P., Shen R., Yang F. 2002.The application of Case Based
ing of the contents, assignment of meanings to at- Reasoning on Q&A System. Proc of the 15th Australian
tributes (as to the phenomenon of event type, ar- Joint Conference on Artificial Intelligence (AI2002), Can-
berra, Australia: 31-37.
gumentation. Henninger S. 2003.Tool Support for Experience-Based Soft-
3 For QA groups: transformation to the node, ex- ware Development Methodologies. Advances in Computers
pansion into QA structure, transformation to the Vol. 59: 29-82.
event net, visualization of a network, analysis of a Hirschman L.& Gaizauskas R. 2001. Natural Language Ques-
condition, choice of a direction of development, tion Answering: The View from Here. Natural Language
Engineering Vol. 7: 67-87.
scrolling of dynamics (on inquiries. Kroll P.& Kruchten Ph. 2003. The Rational Unified Process
4 For text: creation, transformation, grammar Made Easy: A Practitioners Guide to the RUP. Addison-
analysis, semantic analysis, transformation to the Wesley.
semantic graph, visual analysis, supporting of a Yang F., Shen R., Han P. 2003. Adaptive Question and An-
phenomenon of attention. swering Engine Base on Case Based and Reasoning
Technology, Journal of Computer Engineering Vol.29
No.1: 27-28.