Domain Modelling for Technical Documentation Retrieval

Tyrväinen, P.,

Domain Modelling for Technical Documentation Retrieval,

Acta Polytechnica Scandinavica, Mathematics and Computer Science Series No. 64, Helsinki 1994, 1 pp. Published by the Finnish Academy of Technology. ISBN 951-666-406-7. ISSN 0355-2713. UDC 681.327:159.95:519.683.5:681.326.34, 171 p.

Download: PDF (1340 Kbytes)

Keywords:

Ontology, domain modelling, information retrieval, technical documentation, product information management, knowledge representation and transfer, thesaurus construction, hypertext.

ABSTRACT

This work addresses the problem of transferring knowledge of large technical systems from the designers to the end users. The proposed solution is to aid retrieval of technical documents by constructing conceptual models to describe the domains of the documentation. The technical focus is on reducing the human effort needed to construct these domain models. First, different text retrieval approaches and the use of domain modelling in them are analysed. For this purpose, a framework of representations used in text retrieval systems is devised. The analysis covers the information retrieval (IR), the natural language processing (NLP), and the hypertext and related document structure approaches. On the basis of this general analysis, the special features of the technical documentation process, text retrieval, and domain modelling in technical domains in particular are discussed. As a result, a list of requirements for ideal knowledge representations aimed at supporting text retrieval and a general principle guiding domain modelling processes in technical documentation domains are proposed. The realisation of the methods is based on the D&T domain models and the DTM modelling system developed in the Esprit II project SIMPR. The D&T models satisfy the requirements presented earlier and emphasise explicit representation of central domain structures to the users. The DTM modelling system supports the creation of conceptual domain models based on available design information databases, utilises structural homogeneity of the information, and enables integration of the models with various retrieval techniques to map the models with texts. This aproach leads to a solution with which it is possible to create complex hypertext-like structures with minimal human intervention by filtering relevant information from existing design data sources. The approach is validated in an industrial pilot.

Postscript:

This PhD thesis is valuable for the researchers of ontology. Especially, in chapter 7 it

represents a two-level ontology with domain-specific ontology classes (called concept classes) and relation classes in between them. The domain-specific entities in an otology follow the constraints of these classes and the semantics of the relationships and attributes defined for them. This approach has been widely accepted by the research community after publication of this work. Further, it
represents text-reference classes for the ontology classes. These can be considered as generalization of e.g. the approach in topic maps, thus making it possible to simulate their associations with information objects as the set of instances of a text-reference class. Further, it
represents mapping methods for text-reference classes. I.e. represents how to define a method for a text-reference class (attached to a specific ontology class) that makes use of XML document structures, natural language processing and other similar methods for dynamically associate each instance of an ontology class with a set of information objects related to it all with one definition. Further, it
represents a method for semi-automating the creation of a domain-specific ontology (called Domain and Task Model) from the data contained in product data models and other design systems. This means that by

defining 10-50 ontology classes with their attributes, relation classes and text-reference classes it is possible to
import about 1000-5000 ontology instances and their relations from design databases and to
automatically generate 5000-20000 references to information objects for each configuration of a large technical product.

The other parts of the thesis are valuable in positioning the contemporary development of ontology research (DAML, OIL, Topic Maps etc.) with the perspective of related disciplines. Do not re-invent the wheel...

Related publications

Tyrväinen, P. Saarinen, P. Hätönen, K., "Domain Modelling for Technical Documentation Retrieval," in Kangassalo, H. Jaakkola, H. Hori, K. Kitahashi, T. (eds.), Information Modelling and Knowledge Bases IV, IOS Press, The Netherlands, 1993, pp. 388-399.
Tyrväinen, P. Saarinen, P. Hätönen, K., "Domain Modelling for Technical Documentation Retrieval," in Kangassalo, H. Jaakkola, H. Hori, K. Kitahashi, T. (eds.), Information Modelling and Knowledge Bases IV, IOS Press, The Netherlands, 1993, pp. 388-399.
Tyrväinen, P., "Hypertext and Text retrieval," in Hyvönen, E. Seppänen, J. Syrjänen, M. (eds.), STeP-92 New Directions in Artificial Intelligence, Vol. 2 - Symposia, Publications of the Finnish Artificial Intelligence Society (FAIS), 1992, pp. 96-105.
Tyrväinen, P. Saarinen, P. Hätönen, K., "DTM -- Domain Modelling for Technical Documentation Retrieval," in Yesha Y. (ed.), Information and Knowledge Management. CIKM-92, Publication of ISMM, ISBN: 1-880843-03-X, 1992, pp. 509-516.
Tyrväinen, P., "Domain and Task Modelling -- Background, Methodology, and Application," SIMPR Document no: SIMPR-NRC-1992-31.3E, 1992.
Tyrväinen, P., "Feasibility of NLP, Index Term Extraction, and Domain Modelling to Processing and Retrieval of Technical Documentation," Natural Language Text Retrieval Workshop Notes from the Ninth National Conference on Artificial Intelligence (AAAI-91), Anaheim, California, July 15, 1991.
Tyrväinen P. "Measures for Problem-Oriented Information Retrieval Process," SIMPR-NRC-1991-26.16E September 1991.
Hätönen, K. Parpola, P. Rämö, K. Tyrväinen, P., "On the Structures of Technical Documentation," SIMPR document No. SIMPR-NRC-1991-26-22e, September 1991.
Hätönen, K. Parpola, P. Tyrväinen, P., "Semi-Automatic Creation of Domain Models Based on Design Information Input," SIMPR document No. SIMPR-NRC-1991-24-17e, September 1991.
Hätönen, K. Parpola, P. Tyrväinen, P., "Text References: Methods to Link Models with Texts," SIMPR document No. SIMPR-NRC-1991-24-18e, September 1991.