Native language influence detection for forensic authorship analysis: Identifying L1 Persian bloggers

Abstract

This article demonstrates and examines the potential use of interlingual identifiers for forensic authorship analysis and native language influence detection (NLID). The work focuses on the practical applications of native language (L1) identifiers by a human analyst in investigative situations. Using naturally occurring blog posts where the writer self-identifies as a native Persian speaker, a human analyst derived and coded sets of non-native features. Two logistic regression models were built: the first was used to select features to distinguish L1 Persian speakers from L1 English speakers in their English writings, the second developed a feature list to contrast L1 languages that are geographically and linguistically close to Persian. The results clearly demonstrate that interlingual identifiers have the potential to aid in determining the L1 of an anonymous author and can be used by a human analyst in a short forensically realistic example text. This article demonstrates that NLID is possible beyond the more common computational approaches and can form a useful tool in the forensic linguist’s toolbox. This study is not a statistical validation study, instead it demonstrates how a sociolinguistic approach can complement more traditional computational approaches.