An advanced NLP and machine learning-based chatbot system for effective Igbo language communication

Abstract

The rapid advancement of artificial intelligence (AI) and natural language processing (NLP) has transformed human–computer interaction globally, yet indigenous African languages such as Igbo remain underrepresented in conversational systems. This study addresses this gap by developing an advanced NLP and machine learning–based chatbot system for effective Igbo language communication. The methodology involved constructing a large-scale annotated dataset of 750,000 user expressions and responses, incorporating intent categories, sentiment labels, and dialectal inclusivity validated by linguistic experts. Preprocessing combined Igbo-specific tokenization, lemmatization, orthographic normalization, and data augmentation techniques, while model development adopted multilingual BERT (mBERT) optimized with multi-task learning for simultaneous intent recognition and sentiment classification. Training employed stratified sampling, grid-searched hyperparameters, and 5-fold cross-validation to enhance robustness and generalization. Results demonstrated significant performance improvements, with the proposed system achieving 92.6% accuracy, 91.8% precision, 92.3% recall, and an F1-score of 92.0%, outperforming Naïve Bayes and deep neural network baselines. Extrinsic evaluations confirmed strong user satisfaction (4.4/5) and high response relevance (4.6/5), particularly in transactional and service-oriented queries. Discussion of comparative studies indicated that the Igbo chatbot surpassed similar systems for Hausa, Yoruba, and Swahili despite smaller corpora, validating the effectiveness of transformer-based approaches combined with linguistically informed preprocessing. This research provides both technological and socio-cultural contributions by advancing inclusive AI for low-resource languages. The findings demonstrate that transformer-based architectures, when tailored with dialect-sensitive preprocessing and augmentation, can close performance gaps and promote equitable digital communication.