Human and Machine Approaches to Primary Language Identification in Four Accents of Nigerian English

This study examined the effectiveness of human and automated approaches in classifying Nigerian English accents. Building on Muhammad-Gombe et al. (2023) , it compared 25 native speaker linguists and 80 native speaker non-linguists of four Nigerian languages (Hausa, Igbo, Kanuri, and Yoruba) in classifying the native language from Nigerian English accents against 13 UK-based phoneticians familiarised with Nigerian English features and an automated accent classification system, Y-ACCDIST. The findings indicated no significant differences in accuracy between these groups, although native speakers were better at recognizing accents from their own first language. Distinct error patterns emerged across groups, with native speakers often confusing similar first languages, UK phoneticians showing more distributed errors, and Y-ACCDIST occasionally misclassifying Hausa as Igbo. This study highlights that native speakers remain critical in language analysis for Language Analysis for the Determination of Origin/Language Analysis in the Asylum Procedure contexts, especially when supported by trained linguists or automatic systems. It also suggests Y-ACCDIST as a complementary tool for efficient, reliable accent analysis, reinforcing the potential of combining native speaker expertise with computational tools to improve language verification processes.