LEARNING OBJECT ORIENTED DESIGN THROUGH
AUTOMATIC MODEL GENERATION
FROM NATURAL LANGUAGE PROBLEM STATEMENTS
INTRODUCTION
A number n of software (SW)
analysts may produce n different (perhaps all of them correct) solutions from
one specific SW requirement document. This happens because Natural Language
(NL) understanding involves syntactic, semantic and pragmatic issues, (different
backgrounds in persons produce different interpretations) and because of
distinct design experiences. I am proposing a technique (called role posets)
able to be automated and a semi-natural Language (4WL) as a main vehicles of a
methodology to accelerate the production of reliable accords between
different stakeholders. The supporting software tool (GOOAL: Graphic Object
Oriented Analysis Laboratory) automatically produces object models (UML
diagrams) from NL (English) statements. These statements describe the original
software requirements. These models are generated by analyzing sentence by
sentence the intermediate language (4W) version of the original sentence set.
With this methodology and supporting software tool, students of Object Oriented
technology courses can realize which design decisions are being taken by the
system while it iterates through the requirement sentence.
THEORETICAL BACKGROUND
Although it has been proven that
Natural Language processing with holistic objectives is a very complex task
(McDonald 1990)(Hobbs 1992), it is possible to extract sufficient meaning from
NL sentences to produce reliable models. Complexities of language range from
simple synonyms and antonyms to such complex issues as idioms, anaphoric
relations or metaphors. Efforts in this particular area have had some
success in generating static object models using some complex NL requirement
sentences. This research proposes the use of a tool-supported methodology that
helps in the object oriented design teaching-learning process.
PREVIOUS RESEARCH
Some methodologies have followed
the approach to formalize the process to create software products with limited
intentions to automates it. This approach shall cover the complete process
(even to the coding stage). Other approach has a clear aim to use specific
techniques to construct an automated software tool to help the user. Under this
tool-oriented approach, we shall consider the user and the tool developer
points of view.
The user's point of view sees the techniques and methodologies
considering the goals and nature of the user.
The tool developer's point of view considers the general view of
constructing a software tool that proves the proposed methodologies.
We analyze the methodologies and the Software Design tools from the Natural
Language System Requirements Specifications (SDNL).
The first relevant published technique attempting to produce a systematic
procedure to produce design models from NL requirements was Abbot (1983). Abbot
suggested a non automatic methodology that only produces static analysis and
design products obtained by an informal technique requiring high participation
with the of users' for decisions.
Cordes and Carver (1988) presented a methodology by which the English
requirements document is converted into a reliable set of knowledge for use
within the specification process. Although this is a tool oriented effort, its
objective is the production of a set of knowledge for reuse when a new problem
is an extension of a previous one.
Saeki, Horai and Enemoto (1989) published the first effort oriented to a more
comprehensive analyses of the software process. Their research present a
process to derive incrementally a formal specification from an informal
specification written in NL. They suggested the use of simple "verb
patterns" to identify linguistic subjects and objects. Their methodology
highly depend on user interactivity. The technique produces a list of nouns and
verbs. The user decides which nouns become classes and which verbs become
methods.
Cockburn (1992) presented a more detailed analysis of NL elements. He related
relational nouns with objects and adverbs with polymorphisms. He analyses the
roles of objects in a high level abstraction.
Osborne and MacNish (1996) present a methodology whose objective is to
eliminate ambiguity in NL requirements. The problems associated with processing
unrestricted NL are established. The impact of multiple sense problems can be
reduced by the use of a Controlled language (CL). This CL can have several
disadvantages: It can restrict and irritate the user and can be difficult to
learn The proposed CL is called Newspeak.
The Da Silva, Baptista's (1996) thesis suggested an object oriented model whose
value in those days (before UML) was very high. He presents a formal
diagrammatic notation (Object-Z) and a set of rules to transform semiformal
into formal specifications.
Mich and Garigliano (1996) presented a very serious effort to automatically
produce good results. They centered their publication on a software tool
that is based on previous software using NL processing, but with wider
objectives (as translations or summary productions). This effort is improved
with the work of Borstler, Cordes and Carver (2001).
Burg and Van de Riet (1996) presented an interesting example that try to
minimize the participation of the user in the job of extracting classes and
relationships from the text. There has not been a lot of information about
their research, but it seems that it includes a very large lexicon to aid in
semantic model validation and provides a new modeling language to illustrate
results.
Hars and Marchewka (1997) presented one of the first important efforts to
produce dynamic analysis products in 1997. They presented an interesting
classification about the obstacles "within",
"between" and "among" the users of a proposed technique or
tool. Their effort claims to be the first technique (and tool) that considers
the meaning of the concepts involved.
The supporting tool's core approach is a word dictionary (about 23,000
words). These are just single root words and every one includes a concept
category (about 15: event, person, location, etc.). The dictionary also
contains word frequency information which was computed from the number of
occurrences of words in a set in a set of volumes from one year of Time
magazine. The process starts transforming the textual sentences into an
internal representation, then a parser construct an equivalent tree
structure. A separate algorithm resolves complex or composite words.
This technique's tools have the capacity to ask the user to resolve
ambiguities not just at a syntactic level but at a semantic level as well.
Boyd (1999) made a good summary of past efforts in his research, and the talks
interestingly about linguistic metaphors in software design and emphasizes
about analysis prepositions, articles, interjections and so on, in addition
to nouns and verbs. A process of syntactic normalizations is introduced.
This technique produces a more precise sentence through the following two
manual steps: Syntax normalization and semantic exploration. This work is
valuable but does not produce automatic generation of results.
Borstler, Cordes and Carver (2001) presented a product that may be the
most elaborated tool currently developed to resolve this problem. The prototype
accepts well formed NL textual use cases and produces, with minimal user
participation UML static diagrams with Classes, objects and simple
relationships. A Valuable feature is the final traceability supported by
hypertext technology.
Overmyer, Lavoie and Rambow (2001) presented a very complete interactive
methodology and prototype. It is a usable tool to provide linguistic assistance
to produce a subset of UML results. However, the text analysis remains in
good part a manual process. It provides reliable tools for the user to analyze
texts but it does not analyze text itself.
GOALS
The supporting software tool
(GOOAL: Graphic Object Oriented Analysis Laboratory) in this research paper
will automatically produce static and dynamic object models from a NL
statements (in English). These statements will describe the original software
requirements. These models are generated by analyzing sentence-by-sentence the
intermediate language (4W) version of the original sentences set.
With this methodology and supporting software tool, students of Object Oriented
technology courses can realize which design decisions are being taken by the
system while it iterates through the requirement sentence.
CURRENT STATUS AND CONCLUSSIONS
The proposed tool has demonstrated
reliable results with simple sentences and has been validated with different
versions of the original sentences. Manual application of the core tool's
techniques has been applied to complex domains (such chess rules) and showed
promising results.
Novel issues of this proposal are the integrated tool assisted methodology and
the automatic generation of dynamic analysis models with educational
objectives.
OPEN ISSUES
Although the proposed tool
presents promising results using simple sentences, it is intended for use with
more elaborated cases.
The current focus of this research is the automatic decision about classes,
objects, relationships, collaborations and sequences. At the end of this
research the tool must produce more complete sets of models and must be capable
of producing reliable results with more complex sets of sentences. Students of
Object Oriented courses will be exposed to using this tool to compare and
validate results against a control group. It is intended to prove this
methodology as a valuable tool to produce memorable learning experiences.
CURRENT STAGE IN MY PROGRAM OF STUDY
I have completed the required
course normal credits and 16 from the required 30 Phd dissertation credit hours.
It is expected that I can complete the complete credits on summer 2003. I hope
to gain a valuable feedback that can help me to define reliable methods to
validate results of my research and software tool and precise the particular
objectives of the project.
BIBLIOGRAPHIC REFERENCES
(Abbott 1983) Abbott R.
"Program Design by Informal English Descriptions"
Communications of the ACM, 26(11): 882-894, Nov 1983
(Borstler 1992) Borstler, Jurgen, Cordes, David and Carver
Doris.
An Object-Based Requirements Modeling Method.
Journal of the American Society for Inf. Science 43(1):62-71, Jan. 1992
(Boyd 1999) Boyd, Nik. "Using Natural Language in Software
Development"
Journal Of Object Oriented Programming February 1999
http://www.jps.net/nikboyd/papers/rhetoric/road/index.htm
1999
(Burg
1996) Burg J.F.M. and Van De Riet. R.P. "Analyzing Informal Requirements
Specifications: A first Step towards conceptual modeling"
Proceedings of the 2th International workshop on applications
of natural language to information systems, Amsterdam,
The Netherlands, IOS Press 1996
(Cockburn 1992) Cockburn A. "Using Natural Language as a metaphorical
Basis for Object Oriented Modeling and Programming",
IBM Technical Report TR-36.0002, 1992
(Cordes 1988) Cordes A.W., Carber D.L. "Generating a Requirements
Specifications Knowledge Base"
Proceedings of the 1988 ACM 16th. Annual conference on Computer
Science, February 1988
(Da Silva 1990) Da Silva Araujo Joao Baptista Junior M.Sc.
"Metamorphosis: An Integrated Object Oriented Requirements Analysis
and Specification Method"
Lancaster University http://citeseer.nj.nec.com
(Hars 1997) Hars, Alexander and Marchewka, Jack T.
"The Application of Natural Language Processing requirements Analysis"
(Journal of Management Information Systems, March 1997)http://www-ref.usc.edu/~hars/pub
(Hobbs 1992) Hobbs, Jerry R., Douglas E. Appelt, Jhon Bear, Mabry Tyson and
David Magerman.(1992)
Robust Processing of real World Natural-Language Texts. SRI international.
Text Based intelligent systems. Pages 13-34.Lawrence Erlbaum Associates
Publishers.
(McDonald 1992) McDonald, David D. (1992)
Robust Partial-Parsing Through Incremental Multi-Alghoritm Processing.
Text Based intelligent systems. Pages 83-100.Lawrence Erlbaum Associates
Publishers
(Mich 1994) Mich L., Garigliano R. [1994]
(NL-OOPS) "NL-OOPS A Tool for Object Oriented Requirements Analysis"
http://nl-oops.cs.unitn.it/
(Osborne 1996) Osborne, Miles KmacNish, C
"Processing Natural Language Software Requirement Specifications"
(1996)Department of Computer Science, Univ. of York, Heslington
http://citeseer.nj.nec.com/97557.html
(Overmyer 2001) Overmyer S. Lavoie Venoit and Rambow O.
"Conceptual Modeling through Linguistics Analysis Using LIDA"
23rd international conference on Software engineering July 2001
(Saeki 1989) Saeki,M.H.Horai, H.Enemoto "Software Development
Process from Natural Language Specification" Proceedings of the 11th
International conference on SW Engineering IEEE Computer Society
Press 1989
(Umeå 2001) Umeå University project-
http://www.cs.umu.se/~jubo/RECORD.html