Computing Perspective in Text - Computational Linguistics and Talking Robots - page 211

Often the author's circumstances of writing, coded in the STAR-1, differ
from the STAR-0 (Sect. 10.2); the STAR-0 codes the circumstances of a con-
tent's origin as a recording of the agent's current recognition and action (Sect.
10.1). However, if the agent writes down the recognition and action data im-
mediately, the STAR-1 and the STAR-0 are practically the same. In this case,
the perspective of the STAR-1, relating the utterance situation to the STAR-0
of the content, is shortened to the perspective of the STAR-0.
Mapping a STAR-0 = STAR-1 content into language creates an effect of im-
mediacy and authenticity. This is what the author achieves in example 11.1.1:
there is no direct reference to any earlier events or locations of a separate
STAR-0 content. Instead, the use of the present tense leads the reader to view
Porto Praya with the eyes of the author from the deck of the Beagle on Jan.
16th, 1832. The reader can relax and enjoy the author's report because he or
she is neither expected nor even able to derive a response which could reach
the author, in contradistinction to a face-to-face question or request.
In DBS, a text like 11.1.1 is analyzed as a sequence of statement dialogues.
As shown in 10.6.7, each elementary statement dialogue includes the deriva-
tion of a STAR-2 perspective by the hearer/reader (otherwise it would be a
monologue). In 11.1.1, the STAR-2 perspective is not used as the basis for a
subsequent hearer/reader response (in contradistinction to Sect. 10.3). It is not
even used for interpreting the indexicals moi and toi (because there are none
in the text). It only provides a perspective as an integral part of the reader's
understanding, looking back from her or his current STAR-2 circumstances to
the author's STAR-1 = STAR-0 circumstances in the year 1832.
While the computation of the speaker/writer and the hearer/reader perspec-
tives based on the STAR-0, STAR-1, and STAR-2 is completely mechanical,
there is another aspect of understanding which relates to differences in the
background knowledge of different readers (cf. FoCL'99, Sect. 21.5). For ex-
ample, a reader who knows the location of Porto Praya on the globe will have
a better understanding of the sample text than one who does not. Similarly, a
reader who has already experienced a walk under cocoa-nut trees, or traveled
on a sailing ship will have a more complete understanding of the text than one
who has not.
Such differences due to different background knowledge apply to the un-
derstanding not only of statement dialogues, but also of question and request
dialogues. This kind of unsystematic variation may be easily modeled by pro-
viding the Word Banks of different artificial agents with different contents. It
is neither an intrinsic part of nor an obstacle to the language communication
mechanism reconstructed in DBS.