VINST Validation and Integration of Specifications
Background
During 1997-98 Dr. Hercules Dalianis
, ph +46 8 70-568 13 59, was invited to, Information
Sciences Institute/ University of Southern California, by Dr.
Eduard Hovy , to do research and to develop a prototype to integrate
STEP/EXPRESS schemata for the car manufacturing industry.
VINST means Validation and Integration of Specifications. The VINST
tool compares two concepts in two formal models and decides how close they
are semantically The tool can then consequently
automatically integrate the two models.
The VINST tool can also translate the formal
model into Natural Language (English) so that the user can validate the
model.The tool was presented at The
European Conference on Integration in Manufacturing, IiM-98, Gothenburg
October 6-8, 1998.
Formal models and language
The VINST method can be used within a large set
of areas where there is a need to integrate/validate two models. Except
of the STEP/EXPRESS domain, we can also do it for languages in the system
development area as for example UML, SDL, database models, business models,
etc.
Metod
The way the VINST tool is implemented is that the tool searches through
all concepts in both models and compares them with regard to of a set of
parameters as concept names, attributes, sub- and super classes of the
text definition of each concept. Each measurement is weighted and is then
inserted in a formula to compute the semantic proximity of each concept-pair.
Each value is then sorted in decreasing order. The Pilot example studied
was the integration of the Electrotechnical application protocol STEP AP212
and the Automotive design application protocol STEP AP214, containing 352
and 501 concepts respectively. The VINST method gave 25 completely overlapping
concepts and 20 partly overlapping concepts, see
article for details. Some of the semantically close concepts might
not look the same at a first glance, as for example, project and
activity,
but can have exactly the same meaning.
Figure 1. Overview of the VINST tool
Domains
Automotive manufacturing industry, telecom, infocom, finance, etc.
An example: Merging two companies.
When two large companies in the utility domain are to be merged, a large
amount of concepts in the businesses are to be redefined, replaced or kept.
These concepts are available in a large set of sub domain in the utility
domain, for example in power distribution/network, power exchange, billing,
process technique, personnel, etc, etc,
In some of these domains in the merged companies there can be descriptions
in machine readable form as database schemata, network descriptions, billing
systems etc. In these cases it is very easy to let the VINST tool read
these concepts and give the first assessment of which of the concepts of
two companies are semantically close. The results from the VINST tool is
a support for the persons who are assigned to investigate all schemata/concepts
of the merging companies.
Profits
The two companies has for example each 100 concepts within one domain,
this means that we have 100 x 100 = 10.000 possible combinations of concepts
which one or many persons have two investigate. If he/she has to investigate
each concept pair for two minutes then it would take 333 hours = 2 months
to process all. With the VINST tool the whole work would take at most two
days. Of course it would be necessary to adapt the tool to the specific
domain but this work would not take more then two weeks. If one counts
all the domains within these two merging companies the gain in time and
money would be enormous.
An other example: Integration of two formal specifications of telephone
services.
A telecom company sells a telephone system, with a set of predefined services
to a telecom operator. The customer (the telecom operator) wants to customize
a set of services. These new services are interacting in a faulty way with
the predefined ones. The VINST tool performs an analysis of the specifications
and gives a report on which concepts in the specifications are defined
in a similar way.
Platforms
The integration part of the VINST tool is implemented in Perl which is
a string handling language primarily developed for UNIX. Perl is also available
also for Windows and Macintosh OS.
The validation part of the VINST tool is developed in Prolog which
is a logic programming language. Prolog is well suited for natural language
processing. Prolog is also available on the above mentioned environments.
References
H. Dalianis. The VINST approach: Validating and Integrating
STEP AP Schemata Using a Semi Automatic Tool. In N. Mårtensson et
al (Eds). Changing the Ways we Work - Shaping the ICT solutions for the
Next Century, IOS-Press, 1998, pp. 211-220, Proceedings of the Conference
on Integration in Manufacturing, IiM-98, in Gothenburg, Sweden, October
6-8, 1998,Abstract
H. Dalianis & E. Hovy. Integrating STEP
Schemata using Automatic Methods. In the Proceedings of the ECAI-98 workshop
on Applications of ontologies and problem-solving methods, Brighton, England,
pp. 54-66, August 24-25, 1998,Abstract
H. Dalianis, P. Johannesson & A. Hedman.
Validation of STEP/EXPRESS Specifications by Automatic Natural Language
Generation. In the Proceedings of RANLP'97: Recent Advances in Natural
Language Processing, pp. 264 - 269, Tzigov Chark, Bulgaria, September 11-13,
1997, Abstract