Native Dialect Influence Detection (NDID)

differentiating between Mexican and Peninsular L1 Spanish in L2 English

Abstract

The current investigation addresses a vital lacuna in forensic authorship
studies, and more concretely, in Native Language Influence Detection (NLID)
research: narrowing down a speaker’s native dialect instead of only their native
language (L1), which might not be enough when carrying out sociolinguistic profiling
tasks. Native Dialect Influence Detection (NDID), the focus of our study,
can thus greatly aid at the investigative level. We approach this topic by providing
a comprehensive analysis of linguistic features that serve to identify two
non-contact dialects of L1 Spanish (i.e., Mexican and Peninsular varieties) when
dealing with data written in L2 English, which come from Tripadvisor. Our main
aim is to investigate if an author’s L2 features can point to their L1 native dialect,
rather than only to their native language. Findings point to L1 dialectal
transfer of punctuation signs, adjectives of affect, and intensifiers: these linguistic
features, even when expressed in an L2, show a culturally bound use. Additionally,
we implemented an automatic classifier that achieved an accuracy of 69% in
categorizing test data, using only linguistic features that have explanatory power
and can aid linguistic theory. This is key for explainability in the forensic context,
which Native Language Identification (NLI) studies tend to neglect (Kingston
2019). Results show that L1 Spanish dialects can be differentiated by analyzing L2
English text, pointing to NDID as a fertile approach for narrowing down candidate
L1 dialects of a language when analyzing L2 data.