A great deal of the research into legibility has been done in the printing
world. [NaFo] The functions of colour, typeface, font size and so on are
reasonably well understood, although there are some ergonomic differences
associated between screens and the printed page.
On readability, the basic rules of thumb for the following
parameters have been established: [LiCo88]
-
accuracy
-
comprehension
-
speed
There is some understanding of how complexity and style of grammar
affects comprehension. However, we have found that there is little work
so far on the structure of real time conversations. Nor is there a great
deal of work where there are more than two participants.
As our starting point we considered a classification of <#1198#> speech acts<#1198#>
from Searle so that we could start to identify multi party protocols:
He classifies speech into 5 illocutionary acts. These are:
-
Assertive: suggest that someone should do something
e.g. ``assert'', ``claim'', ``argue'', etc
-
Commissive: state that we will do something.
e.g. ``commit'', ``promise'', ``threaten'', etc
-
Directive: order someone to do something
e.g. ``request'', ``order'', ``solicit'', etc
-
Declarative: Naming/Re-assigning - such and such is the case.
e.g. ``declare'', ``name'', ``abbreviate'', etc
-
Expressive: state feelings about something
e.g. ``thank'', ``apologise'', ``greet'', etc
We can use these acts to help structure the exchanges in a conference. A
directive for example should always require an answer, agreement or
acknowledgment and the conferencing system will make provisions so that the
person to whom the directive is made will be forced to make a reply before
they can continue with the conference. An assertive will normally beg a
question (directive), an argument (an assertive not logically consistent
with the last made assertive) or evidence or opinion backing up the previous
statement (an assertive or expressive). Breaking down communications into these
categories enables us to semi-structure (pre-compile into the conferencing
system) conversations in advance, enabling the users to be guided and
understood more efficiently in their communications. [CaDi89]
Even more importantly though placing statements under particular speech act
headings will remove a significant amount of ambiguities as well as enabling
the hearer a quicker route to understanding the speakers intentions. The
proposition ``Close the door'' for example could be a promise or threat, but
equally could be a request or order. Placing all propositions under their
speech act heading will enable the hearer to remove many of the ambiguities
that arise through the loss of information which happens when a speaker's
tone and expression are unavailable.
Not only can we use these acts to give more structure to the exchanges
of text between users, but also for giving weight to their <#1201#> bids<#1201#>
for the floor. This has been used by other workers in the area,
although only for off-line systems rather than real time ones.
Mechanical protocols generally come from a very limited range of the
set possible. E.g. ARP ``who-has'', or RPC ``request-response'' or the
TELNET IACs, ``do, don't, will, won't''. We believe that the use of speech
acts allows a much wider expression for exchanges without losing
structure altogether. [ARP82][RPC81][Tel83]
The Coordinator is a system which, by exchanging messages restricted
to some types chosen to express some of the same ideas as speech acts,
(though not in real time) allows better time management of actions.
[Wino88]