342 JOURNAL OF LAW AND POLICY
F. ReplicabilityThe protocol developed through repeated validation testing
must be repeatable by others who use it. Methods within the
protocol also must be repeatable through the implementation in
computer software or the strict operationalization of terms and
procedures. Implementing a method in computer software is a
sure way of providing objectivity and maintaining consistency.
Systems can be designed so that each user can tweak
parameters, thereby changing the algorithm. However, these
tweaks might not be visible later. Accordingly, such systems do
not maintain consistency in running a method, and the fact that a
method is implemented in software does not necessarily
guarantee that it is completely replicable.
G. The Method’s Relationship to Academic and Industrial
UsesFinally, the research environment should be related to
academia and/or industry by the sharing of knowledge,
techniques, methods, or software. The researcher forensic
linguist is part of a larger community of computational linguists,
psycholinguists, corpus linguists, theoretical linguists, and
computer scientists where forensic applications are just one
application of common techniques, methods, and software put
together in novel ways. For instance, text classification
techniques were originally designed as part of summarization
schemes but later became useful for finding plagiarism and
duplicates within large electronic collections, just as DNA
testing was originally used for paternity before it was applied
forensically.
Forensic author identification methods should relate, in some
recognizable way, to a theory of language, since the method is
seeking to identify authorship based on language (rather than
handwriting, ink, or IP address). Linguistics obviously offers the
fullest theories of language, with the generative theory being the
best developed. The generative theory of language includes