Computing Perspective in Dialogue - Computational Linguistics and Talking Robots - page 192

The indexical ça is defined to point at the value of the 3rd attribute of the
STAR-0. In contradistinction to 10.1.5, the reference to Fido is not by name
here, but by indexical.
For STAR-0 contents, the attributes R and 3rd are optional in that their val-
ues need only be defined if the content contains toi and ça pointers, respec-
tively. For STAR-1 contents (speak mode) and STAR-2 contents (hear mode),
a value for the R attribute is obligatory; in small children, the R value may be
viewed as fixed to the mother.
In summary, values for S, T, and A attributes in the STAR are obligatory for
all anchored contents. Thereby a rough idea of the location, of the date and
the time of day, and of oneself will usually suffice for all practical purposes.
Loss of the STA parameter values for any length of time, however, results in
complete disorientation of the agent.

10.2 Speaker's STAR-1 Perspective on Stored Content

The next day (T: 1930-07-04), Simone and J.-P. meet in the Café de Flore on
the Left Bank. To get the conversation going, J.-P. begins a statement dialogue
by uttering the following declarative sentence to her (in French):

10.2.1 STAR-1 EXPRESSION WITH 1 ST AND 2 ND PERSON INDEXICALS
I wrote you a letter yesterday. STAR − 1
At this point, J.-P. is dealing with two STARs, (i) the STAR-0 defined in 10.1.3
and used in the anchored content 10.1.4 and (ii) the STAR-1 for the utterance
indicated in 10.2.1. The S, A, and R values of these two STARs in J.-P.'s mind
happen to be the same, 7 but the T values differ. This difference constitutes a
second kind of perspective: J.-P. is looking back onto a content created in his
mind in the recent past. The English surface of 10.2.1 reflects this perspective
by means of (i) the past tense form of the finite verb and (ii) the adverbial
modifier yesterday .
At the level of content, the automatic coding of the speaker perspective is
based on the following DBS inference, called STAR-1.1. It takes an anchored
STAR-0 and the speaker's current STAR-1 content like 10.1.4 as input:

10.2.2 STAR-1.1 INFERENCE FOR TEMPORAL SPECIFICATION

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦ ⇒

⎡
⎣

⎤
⎦

S: L
T: D
A: N
R: O

S: L
T: D
A: N
R: O

S: L
T: D
A: N
R: O

⎡
⎣

⎤
⎦

⎡
⎣

⎤
⎦

verb: α
sem: β
mdr: γ

verb: α
prn: K

adj: γ
mdd: α
prn: K+M