This status report describes what has been achieved during the first period of
the VOLVEX Project, July 96 - Aug 97. In respect to interviews and
implementation, meetings with our industrial contacts at VOLVO Data, what
dissemination activities that have taken place and finally what the objectives
are for the future work.
VOLVEX is a three and half year project and one year has passed.
Contact was taken with Per Brorson and Stefan Lindahl at Volvo Data and a
number of meetings was held in Gothenburg to discuss Volvo's problems with
implementing the STEP/EXPRESS standard within the five Volvo companies. Volvo
Data, Volvo Cars, Volvo Trucks, Volvo Penta and Volvo Aero. The project at
Volvo dealing with this is called the PDMI project.
From VOLVO we obtained a PDM (Product Data Managment) model in EXPRESS (textual
form) and EXPRESS-G (graphical form), expressing "Part". Per Brorson and Stefan
Lindahl at VOLVO Data told us that they had difficulties to understand and
explain the "Part" model to personnel at Volvo and asked us to investigate the
problem.
We investigated the Part-model and designed a questionnaire asking questions
about the model where the respondents should answer in Natural Language English
how to explain the model in an appropriate way. The questionnaire contained the
"Part"-model in graphical EXPRESS-G, in textual EXPRESS and also a dictionary
to explain the various used concepts in the model.
The questionnaire was handed out to personnel at Volvo and also to researchers
at DSV-KTH. The questionnaires were collected from Volvo and DSV-KTH and
analysed.
An important finding was that people relied heavily on dictionary information
to describe the model although the dictionary information was not contained in
the model. Furthermore, people experience a need to use instances to
exemplify/illustrate their different answers. The rest of the findings was as
expected; the use of aggregation in the natural language texts, pronoun or
referring expressions.
Regarding the use of concepts from AP214 Core Data for Automotive Design
Process in the "part"-model, we found that very few concepts were used.
The implementation work proceeds as in the work plan. We have used a natural
language generation system called ASTROGEN - Aggregated deep and Surface naTuRal language GENerator
which is
written in Prolog. A domain lexicon has been implemented based on the answers
from the questionnaires.
The ASTROGEN generator has been extended with a number of new sentence planning
features as clause delimitation, use of dictionary information so called canned
text and a simple content selection mechanism. (All the details are desc
Some tools for EXPRESS model treatment have been investigated as well as
natural language generation tools. EPM Technologies EDM (EXPRESS Data Manager)
has been considered for the translation from EXPRESS to Prolog syntax (This
part is today simulated by hand). One possibility is to write the translator in
Prolog. The FUF (Functional Unification Formalism) has been tested for the
natural generation, however it was to slow and inflexible therefore we
developed our ASTROGEN generator for the purpose of generation of natural
language from EXPRESS specifications.
The VOLVEX project was presented in Brussels Dec 4, 1996 for AIT-Advanced
Information Technology in Aerospace and Automotive Industry. The
representantives came from Mercedes, Rover, Fiat and Dassault Aviation.
The paper Validation of STEP/EXPRESS Specifications by Automatic Natural
Language Generation was presentated at to RANLP-97. The
Second International Conference, Recent Advances in Natural Language Processing
11-13 September 1997, pp. 264 - 269, Tzigov Chark, Bulgaria. The paper describes the work carried out in the VOLVEX project.
http://www.dsv.su.se/~hercules/papers/STEP-NLGvalidation-foto.html)
H.Dalianis, P. Johannesson & A. Hedman:Validation of STEP/EXPRESS Specifications by Automatic Natural Language Generation. In Proceedings of RANLP'97: Recent Advances in Natural Language Processing, pp. 264 - 269, Tzigov Chark, Bulgaria, September 11-13, 1997,
H. Dalianis & P. Johannesson. Explaining Conceptual Models -- An Architecture and Design Principles. In David W. Embley and Robert C. Goldstein, (Eds.), ER'97, 16th International Conference on Conceptual Modeling, Lecture Notes in Computer Science No. 1331, pp. 215-228, Springer Verlag, 1997
Abstract
H. Dalianis. Natural Language Aggregation and Clarification using Cue Words. In Proceedings of the 6th European Workshop on Natural Language Generation, pp. 6-16, March 24 - 26, 1997, Gerhard-Mercator University, Duisburg, Germany,
Abstract
H. Dalianis & E. Hovy. On Lexical Aggregation and Ordering. In Proceedings of the 6th European Workshop on Natural Language Generation, pp. 17-27, March 24 - 26, 1997, Gerhard-Mercator University, Duisburg, Germany,
Abstract
H. Dalianis & F. Persson. Reuse of an Ontology in an Electrical Distribution Network Domain. Presented at the AAAI 1997 Spring Symposium Series, Ontological Engineering, pp. 25-32, March 24 - 26, 1997, Stanford University, California.
Abstract
Due to the budget cut down we did not carry out a thorough investigation of the
AP214 as stated in the original project plan. We have for example only studied
one EXPRESS specification which is based on a very small part of AP 214.
One of the spin off effects of the VOLVEX project has been our work with an
ESPRIT project proposal called VINST (Validation, Schema
Integration and Explanation of STEP/EXPRESS specifications.) The
plan is to use the technique and competence developed in VOLVEX project for the
VINST project. Two Swedish companies are involved in this proposal Scania AB
and Bris Data. The VINST ESPRIT Proposal will be submitted to the European
Commission in Dec 1997.
Design the text planner, i.e. decide on how the discourses should look like in
the generated text.
Implement the text planner. Are there implication on the sentence planner?
Study the interaction between the text and sentence planner.
Study UML and EXPRESS-2 regarding the second generation STEP languages and
their applicability for natural language generation.
Putting together a stable prototype of all components. Testing the prototype to
find out what is missing. Demonstrate prototype to design departments at car
companies to obtain comments.
Study ontologies as a mean to support the aggregation process in the sentence
planner as well as the lexicon construction. Study the problem of sentence
delimitation, i.e. how long should a sentence be?
Adding comments from users to the prototype and further develope the prototype.
Second part of research project. Investigate other AP's to see if the technique
of generating natural language is applicable. Which parts of the interviews can
be removed and is it possible to use the text and sentence planner, the surface
grammar and part of the lexicon in the new domain (AP) ?
Writing scientific papers for publication of the results from 1 July 98 - 30
September 98. Carry out a study of all tools around STEP/EXPRESS to see if our
generation tool will fit in.
Write guidelines for how to create lexicons and to (re)use our natural language
generation tool in other domains (APs).
Is there any parser translating from natural language to EXPRESS and if there
is one, what is needed to add a generation component to it ?
What will be needed to create a natural language to EXPRESS tool?
This period is for all the delayed work adding up the final comments and doing
the final literature search. Writing scientific papers for publication.
Writing final report of project
Some of planned work might be carried out in an other order.
* Handbook (Guidelines for how to create lexicons and to (re)use our natural
language generation tool in other domains (APs).
* A natural language generation prototype validating the results of the
project.
* Scientific papers
Project Leader and Applicant
Tekn Dr Hercules Dalianis (PhD)
ph 08-16 49 16
mobile 070-568 13 59
email: hercules@dsv.su.se
http://www.dsv.su.se/~hercules/