Developing an Original Methodology for Computational Analysis of African Oral Literary Corpora

Computational literary analysis methods including corpus linguistics, stylometric analysis, and machine learning-assisted text classification have been applied extensively to European and American literary corpora but have barely been adapted for the distinctive challenges of African oral literary data, including tonal language transcription, formulaic repetition structures, performative variability, and multilingual code-switching. This study developed an original methodology for computational analysis of African oral literary corpora, with Yoruba and Igbo oral poetry as primary test cases. A methodology development approach was employed across three phases. Phase one conducted a systematic review of existing computational approaches to oral literary analysis, identifying twelve studies and their methodological limitations in relation to African oral corpus properties. Phase two conducted empirical testing of four candidate computational approaches (n-gram analysis, topic modelling, stylometric analysis, and social network analysis of character relationships) on a pilot corpus of 150 Yoruba oriki and 120 Igbo uli poem transcriptions. Phase three developed the original Tonal Corpus Analysis Methodology (TCAM), which provides adapted protocols for tonal language literary data, formula detection appropriate for oral compositional structures, and comparative analysis across performer versions of the same oral text. The TCAM was validated through application to an independent corpus of Hausa oral heroic poetry at Bayero University Kano. Expert review by 15 computational literary scholars and African oral literature specialists confirmed the methodology's originality and rigour. The study recommends TCAM as the standard computational approach for African oral literary corpus analysis in the emerging field of Nigerian digital humanities.