Information Technology Reference
In-Depth Information
dialog topic vector
. The several components in this vector provide probabilities for each
predefined Wikipedia category possessing a relation to the considered concept terms.
If a probability again exceeds a given probability threshold, its corresponding category
constitutes the current topic of the ongoing dialog. In case the described conditions are
fulfilled several times within one topic tracking process, the system is not able to de-
termine one single Wikipedia category to be the current dialog topic but rather keeps
all topic options open. Otherwise, that is if one dialog topic could be identified, the un-
derlying dialog topic vector is included in the next identification step to keep track of
this dialog topic subsequently. For this purpose, it is treated as a concept topic vector of
the current utterance and is thus compared to all concept topic vectors of the following
utterance to search for topical overlaps. Figure 2 graphically presents possible results
of the topic tracking process for our example dialog introduced in 2.1 by means of a bar
diagram. As reaching a probability
0.5 after scaling and thus exceeding the thresh-
old represented by the horizontal line in black, the categories
“Regions”
and
“Sports”
constitute the dialog topics within this illustration.
Utterances which do not provide any concept information, like the utterance
“I
know.”
, have no impact on the probabilities for the several dialog topics.
≥
Topic Shift Detection.
As mentioned before, we distinguish between a
topic leap
as
described by Svennevig (1999) and a
topic drift
as introduced by Hobbs (1990). Based
on this, systems are capable of detecting radical topic shifts enabling the particular
conversational agents to generate an appropriate conversation behavior. According to
this, the agent might refer to this topic leap via a suitable utterance such as
“What made
you think of this topic?”
.
To distinguish between the two types of topic shift automatically, the transition from
one dialog topic to the next is evaluated based on the outcomes of the topic tracking
process. That is, if no topical overlap between the utterances
utt
1
and
utt
2
can be
determined, the system detects a topic leap. In contrast, a topic drift is characterized
in that topical overlaps to both the old and the new dialog topic exist during the topic
transition as shown in Figure 2.
Topic Labeling.
To be able to refer to a dialog topic later on, for example in another
dialog, a descriptive topic label has to be defined. Wikipedia provides topic labels in
terms of category titles. Thus, a topic can be labeled with the title of the Wikipedia
category that constitutes the current dialog topic. Thereby, the labels do not have to be
mentioned during dialog before as they are already existent. However, some category
titles might need to be changed to more intuitive labels. The category title “Leisure”,
for instance, can be replaced by “Hobbies” as the latter provides a more humanlike term
for a topic raised in smalltalk conversations.
4
Making Artificial Agents More Topic Aware
So far, we described how to detect topics in ongoing dialog automatically by means
of collaborative knowledge provided by Wikipedia. However, to emulate humanlike
