Engineering Analysis and Recognition of Nigerian English: An Insight into Low Resource Languages

A comparative analysis between Nigerian English (NE) and American English (AE) is presented in this article. The study is aimed at highlighting differences in the speech parameters, and how they influence speech processing and automatic speech recognition (ASR). The UILSpeech corpus of Nigerian-Accented English isolated word recordings, read speech utterances, and video recordings are used as a reference for Nigerian English. The corpus captures the linguistic diversity of Nigeria with data collected from native speakers of Hausa, Igbo, and Yoruba languages. The UILSpeech corpus is intended to provide a unique opportunity for application and expansion of speech processing techniques to a limited resource language dialect. The acoustic-phonetic differences between American English (AE) and Nigerian English (NE) are studied in terms of pronunciation variations, vowel locations in the formant space, mean fundamental frequency, and phone model distances in the acoustic space, as well as through visual speech analysis of the speakers’ articulators. A strong impact of the AE–NE acoustic mismatch on ASR is observed. A combination of model adaptation and extension of the AE lexicon for newly established NE pronunciation variants is shown to substantially improve performance of the AE-trained ASR system in the new NE task. This study is a part of the pioneering efforts towards incorporating speech technology in Nigerian English and is intended to provide a development basis for other low resource language dialects and languages.