A knowledge-based document layout system now in use in the production
of the British Telecom Yellow Pages
Overview
Formation is a knowledge-based document layout system, originally designed
for large classified
directories and catalogues. It consists of a graphical user interface
to a batch layout engine. The heart of this engine is an interpreter for the
Layout Style Specification Language (LSSL), in which document layout styles
are programmed. The Formation
GUI offers dialog-based modification and reconfiguration of programmed styles, along with
graphical display and control facilities for laying out individual documents.
Business Perspective
Pindar Set Ltd., the UK company that originates the British Telecom Yellow Pages,
has developed an international business strategy to market generic, customisable,
knowledge-based layout software and services to publishers of classified directories and
catalogues. Pindar's clients, such as BTYP, increasingly demand faster and more flexible
responses to requests for change in the layout of their pages, in order to improve
their correctness, useability, and aesthetics. Pindar also wishes to attract
new clients from around the world to its directory and catalogue services.
Formation
is allowing Pindar to implement this strategy, providing the company with a system
offering much greater flexibility and higher throughput than the many other batch layout
systems in regular use around the world.
In addition to the software itself, Pindar required a full understanding of the system
so that they could maintain it themselves in future, and develop it further independently
of external support. As a technology transfer organisation, AIAI was able to provide
visiting worker facilities through which Pindar staff could work with, and learn from,
experienced members of the Formation
project team.
Technical Perspective
Formation is a knowledge-based layout system, in the sense that it operates
using the same kind of expert knowledge of principles and procedures that are
employed by a human layout artist doing the same task. In contrast to other automatic
layout systems, it does not do a constraint-based search for some nominal 'best'
solution out of many possibilities. Rather, it uses the principle 'follow correct and
explicit procedures and rules and you'll get the one and only acceptable result'.
Layout styles are expressed in LSSL, an object-oriented, domain-specific language
with primitives such as get-space, align, paste and
unpaste. Each style defines:
- the page geometry, e.g. the number of columns and vertical resolution
within a column;
- the types and properties of items placed on the page,
e.g. heading, listing, advertisement;
- a single overall placement strategy, e.g. place advertisements before
listings, prioritising the re-placement of any item unpasted to make space for another;
- any number of methods which code, in a highly modular fashion,
the various aspects of the layout, e.g. 'keep listings in alphabetic order',
'place larger ads closer to the bottom of the page', 'insert continuations where
a section spans several pages', and so on.
The strategy and methods can be parameterized
to make them flexible and reusable in different styles, and the Formation
GUI allows the non-programming operator to combine and configure methods quickly and
easily between layout runs.
Formation can lay out a typical classified telephone
directory at speeds in excess of 1500 pages per hour on a typical Pentium PC.
Last updated 29th May 1997
by Andrew Casson-du Mont