Information Technology Reference
In-Depth Information
behaviour of police officers and suspects, we describe several fragments from our
corpus in terms of these concepts (Sect.
16.5
). We conclude with our thoughts on the
creation of a computational model for our virtual agents based on the combination
of models and theories (Sect.
16.6
).
16.2
Corpus Analysis
In this section, we outline how we analysed the behaviours of police officers and
interviewees in a corpus of police interviews. We look at the behaviour of suspects
and police officers because we feel modelling the interaction between both parties
is necessary to create a believable virtual suspect. In Fig.
16.1
, we show the steps
we took in our analysis.
We started with a corpus of police interviews (step 1), the
Dutch Police Interview
Training Corpus
(DPIT Corpus), see Sect.
16.2.1
. From this corpus, six observers
independently selected fragments that they thought were “interesting” in some
way (2). For example, these were fragments in which a change in mood or
atmosphere took place or fragments in which behaviour could be observed that
1.
2.
3.
4.
5.
Select
Fragments
Annotate
Fragments
Create
Questions
Rate
Fragments
DPIT Corpus
6.
Factor
Analysis
13.
12.
Future Work:
Computational
Model
Evaluate
Meta-theory
7.
Interpret
Factors
11.
10.
9.
8.
Match
Factors and
Concepts
Concepts
from
Theory
Meta-
theory
Select
Theories
Fig. 16.1
Diagram showing the steps taken towards a computational model of police interviews
Search WWH ::
Custom Search