TXT 4N6 477
In conclusion, the aim in the Birks analysis was to provide a
sound evidentiary analysis based on a pairwise discrimination of
Amanda and Christopher Birks’ respective vocabulary choices in
their text messages. My method seeks to avoid, as much as
possible, my own biases affecting the selection of features, the
analysis, or the drawing of conclusions. In addition, in the
analysis presented in this article, I shall extend the actual
approach taken in the case to demonstrate a statistical approach
that can further support future analyses.
II. METHODS
This section follows with a description of the available data
and the analytic approach taken in the Birks analysis.
Justifications are provided for the decisions taken.
A. DataThe text message data were provided by Staffordshire Police
as part of their investigation into the death of Amanda Birks,
and I was a paid expert witness for the prosecution.^38 Case
details and the forensic analysis were released to the UK
media.^39
Two small corpora of text messages were provided in the
form of spreadsheets identifying phone numbers, names of
recipients, sending times, and text for each SMS message. The
first spreadsheet contained collated text messages sent by
Amanda Birks (“AB”) to eleven separate recipients over a
period of several days prior to January 17, 2009, the date of her
death. As AB’s phone was destroyed in the house fire, police
collected this information from reports produced by forensic
telecommunication engineers who downloaded the messages
from the recipients’ telephones.^40 After a number of automated
(^38) Subsequent to the conclusion of the case, permission was obtained
from the police to use these materials in teaching and research.
(^39) See, e.g., Businessman Admits Murdering Wife, supra note 1.
(^40) I was also provided with the raw telecommunications engineers’ forensic
reports, but I used these only to check the accuracy of the spreadsheets.