Information Technology Reference
In-Depth Information
Problem:
COHERENCE: Which component / who controls the
sentence cited
sentence sub-
ject
to perform this action?
Additionally to the two above question types, based completely on the inference
rules presented in Section 2.1, we address a further problem, resulting from passive
sentences: whenever we come across a passive sentence, we check if the subject of
the sentence coincides with the receiver of the last message. If this is not the case, we
generate the following question:
Problem:
Problematic sentence: passive, actor unspecified (Who / which com-
ponent controls the
sentence cited
sentence subject
?)
As an example, the set of questions addressing the problems of the scenario on page 95
is presented in Table 3.
Ta b l e 3 .
Questions referring to the specification text, generated for the scenario on page 95
1. The instrument cluster is turned on
Problem:
Problematic sentence: passive, actor unspecified (Who / which component
controls the instrument cluster?)
2. The instrument cluster stays active for 30 seconds
(a)
Problem:
Problematic active sentence: object ”seconds” is not contained in the list
of actors. Message receiver cannot be identified.
(b)
Problem:
COHERENCE: Which component / who controls the instrument cluster
to perform this action?
For the questions referring to the model, we further develop the idea shown in Fig-
ure 2: there, the inferred missing message was represented as a “?”-marked message. To
generate the same types of questions as above, we introduce a special actor named “???”
and send all the messages where the message receiver is not explicitly specified, to the
“???”-actor. Similarly, every missing message is “?”-marked. Missing messages in the
sense of Section 2.1 are split into two: a message from the inferred message sender to
the “???”-actor, and a message from the “???”-actor to the inferred receiver.
The set of questions generated for the example on page 95 is shown in Figure 3. It is
easy to see the correspondence to the questions referring to the specification: Question 1
from Table 3 is represented by two “?”-messages before the first assertions, Question 2a
is represented by two messages between the assertions, and Question 2b by the “?”-
messages after the second assertion.
3.2
Feedback Generation for Automata
The algorithm for automata generation presented in Section 2.2 infers, if necessary, the
source state of the transition from the discourse context. It is easy to turn this inference
