The IDAS project investigated how natural language generation and
hypertext techniques can be used in a novel way of producing technical
documentation. Traditional technical documentation suffers from
a number of problems:
IDAS is a novel kind of online documentation system based on the
following key components:
- It is difficult to reuse previous documentation for similar products.
- It is difficult to produce documentation in multiple languages.
- Documentation is bulky and complex to use.
- Documentation is insensitive to user needs.
- The same information may be repeated in many different places, with no
control over consistency.
In IDAS, text is derived from a single non-textual documentation
database, rather than text being itself the primary representation used.
This makes documentation adaptable and reusable.
- An AI knowledge representation language for representing
the object being documented.
- A set of schemas for retrieving information from this
``documentation database'', given a documentation query and information
about the user.
- A natural language generation system which converts this information
into text appropriate to the abilities and interests of the reader.
- A hypertext delivery system for presenting this text to the reader
as interactive hypertext allowing the selection of followup questions.
The initial IDAS prototypes were for users of Automatic Test Equipment.
We were working in collaboration with Racal Research Ltd,
Racal Instruments Ltd and Inference Europe Ltd.
For more information, see the following documents (though the PostScript
versions available here are not identical to the published versions):
E. Reiter, C. Mellish and J. Levine,
``Automatic Generation of On-Line Documentation in the IDAS Project'',
Procs of the 3rd Conference on Applied Natural Language Processing,
Trento, Italy, April 1992.
E. Reiter and C. Mellish,
"Optimizing the Costs and Benefits of Natural
Language Generation", in
Proceedings of IJCAI-93, pp1164-1169, Chambery, France, published
by Morgan Kaufmann, 1993.
Reiter. C. Mellish and J. Levine, ``Automatic Generation of Technical
Applied Artificial Intelligence Vol 9 No 3, pp259-287, 1995.