Developing a framework for the explanation of interlingual features for native and other language influence detection.

Abstract

This article demonstrates the benefit of taking an explanation-based approach in the development of features for computationally supported systems used for linguistic analysis in forensic contexts. As a focal point it considers Other Language Influence Detection (OLID) as well as its related field of Native Language Identifcation (NLI). An explanation-based approach allows the forensic linguist to understand the implications of the presence or absence of features as they vary across the contexts and situations s/he might encounter. The authors present a qualitative framework for types of explanation and show how different types of explanations are needed to develop a full and rich language-influence feature set. The authors are not advocating a strict or inflexible typology of feature explanation but are seeking a richness of explanation at a variety of levels of analysis instead. This in turn can be developed into computational approaches, which the authors contend will therefore be stronger and more applicable to forensic casework contexts.

Publication DOI: https://doi.org/10.21747/21833745/lanlaw/6_2a2
Divisions: ?? 53981500Jl ??
College of Business and Social Sciences > Aston Institute for Forensic Linguistics
College of Business and Social Sciences > School of Social Sciences & Humanities
Additional Information: Este trabalho está licenciado com uma Licença Creative Commons - Atribuição-NãoComercial 4.0 Internacional.
Publication ISSN: 2183-3745
Last Modified: 30 Sep 2024 12:20
Date Deposited: 28 Jan 2020 13:02
Full Text Link:
Related URLs: http://ojs.letr ... ticle/view/6674 (Publisher URL)
PURE Output Type: Article
Published Date: 2019-12-31
Accepted Date: 2019-12-01
Authors: Kredens, Krzysztof (ORCID Profile 0000-0001-7038-9478)
Perkins, Ria (ORCID Profile 0000-0001-6193-1456)
Grant, Tim (ORCID Profile 0000-0002-5155-8413)

Download

Export / Share Citation


Statistics

Additional statistics for this record