How does PLINTH solve these problems?
Firstly, PLINTH
augments hypertext networks with semantic properties. The
author can assign types and slots to nodes and links to mark the function of
the nodes in the network and the structural, logical and rhetorical relations
between them. This allows the author to:
- create, manipulate and browse design rationale (DR) using
hypertext networks based on typed node-and-link based DR models like IBIS
and DRL;
- connect document nodes to their underlying DR structures with ordinary
hypertext links;
- show multiple structural, logical and rhetorical views of a
single document by activating or suppressing different sets of node and link
types, thus reducing the perceived complexity of the network and the reader's
cognitive overhead of understanding it.
PLINTH then supports this augmented hypertext with rule-based intelligent
navigation, to interactively compute a customised path through the document
for the reader, based on the semantic properties and directed by commands and
queries. This reduces the cognitive overhead problem even further. The
author can write sets of navigation rules to perform different text retrieval
or consultation functions for each view of the network. Here are some
examples:
- free browsing:
- Here the navigator follows basic structural links like
next
and part-of
in response to commands from the reader like
next, previous, up, back, and so on, displaying
the text of each node visited. Where a DR node is attached to a document node,
the navigator can dip into the DR network and browsing rules for exploring
this will be activated.
- consultation:
- The consultation facilities of conventional
expert systems for regulations are achieved using navigation based on logical
typing of nodes and links, e.g.
requirement
,
scope
, necessary
, sufficient
,
applies-to
and so on. In this case the navigator follows the
links in logical order, assigning truth values to nodes by asking the reader
whether or not the conditions expressed in the node text are met. These truth
values are combined according to the link types in order to decide which
requirements apply and advise the reader on how to satisfy them. This method
alleviates the expressiveness, parsimony and maintenance
problems:
- The texts of individual clauses need not be expressed as rules. There is
no expressiveness problem with high-level functional concepts like
`adequate' and `sufficient' which are difficult to represent, neither is there
a parsimony problem of having to formalise clear, concise statements
which do not benefit from it. However, it is sometimes useful to be
able to break down clauses further, e.g. to do a calculation expressed
in the text as a table or formula. For this purpose, PLINTH allows local
rules to be attached to nodes and activated when the node is visited;
- Different documents with the same node and link types can use the same
set of basic consultation rules (augmenting them with local rules where
necessary), thus reducing the size and complexity of the rule-base and the
problem of maintenance. Once those rules are written, an author without
programming skills can write a new document and `program' the required logical
breakdown simply by assigning types to the nodes and inserting links between
them.
- design rationale processing:
- The process of building a DR network can be supported by navigation rules
which not only traverse existing parts of the network but also allow the
author to extend it in a principled manner.
What does PLINTH provide?