Development and Evaluation of An Explainable AI-Based Chatbot for First-Aid Assistance in Domestic Accident Scenarios among Ebonyi State Indigenes
IOSR Journal of Computer Engineering · 2026-05 · 28 (5) · 48-58
DOI: 10.9790/0661-2805034858 ↗Kurz gefasst
Fachartikel in der Zeitschrift IOSR Journal of Computer Engineering, 2026-05. Verfasst von Jeremiah und 3 weiteren Autor:innen.
Abstract (Originalsprache)
Domestic accidents constitute an important source of preventable injury and may require immediate first-aid intervention before professional medical care is available. Artificial intelligence (AI)-enabled conversational systems offer a potential mechanism for delivering accessible first-aid guidance; however, their reliability, transparency, and ability to correctly identify accident scenarios remain important concerns. This study developed and evaluated an explainable AI-based chatbot for first-aid assistance in domestic accident scenarios, with the long-term goal of supporting accessible first-aid decision support among populations in Ebonyi State, Nigeria. A natural language processing and machine-learning framework was developed for first-aid intent classification. A labelled first-aid dataset comprising 93 user utterances across 21 injury and emergency-related classes was used for the preliminary experiment. Text was normalized and represented using Term Frequency–Inverse Document Frequency (TF-IDF) features with unigram and bigram representations. Four classifiers—Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), and Multinomial Naïve Bayes (NB)—were evaluated using four-fold stratified cross-validation. Performance was assessed using accuracy, precision, recall, and F1- score, while a separate stratified hold-out set was used to examine the generalization performance of the bestperforming classifier. Explainability and emergency-escalation mechanisms were incorporated into the proposed chatbot architecture. SVM produced the strongest overall performance, achieving 85.96% accuracy, 80.95% precision, 86.31% recall, and 82.10% F1-score. LR achieved 66.71% accuracy, 61.31% precision, 66.07% recall, and 61.87% F1-score; RF achieved 59.19%, 52.29%, 59.52%, and 53.04%, respectively; while NB achieved 36.55%, 28.26%, 36.31%, and 29.95%. SVM achieved 83.33% accuracy on the independent hold-out evaluation. The findings demonstrate the feasibility of machine-learning-based first-aid intent classification as a foundation for an explainable conversational first-aid support system. SVM demonstrated superior performance among the evaluated models. Nevertheless, the preliminary dataset size and class distribution limit generalizability. Future research should incorporate larger, expert-annotated and culturally representative Nigerian datasets, transformer-based models, clinical expert validation, explanation-quality assessment, and prospective user-safety evaluation before real-world deployment.
Zitieren
APA
Jeremiah, Chukwu, Nneka, A. J., & Chukwuma, N. (2026). Development and Evaluation of An Explainable AI-Based Chatbot for First-Aid Assistance in Domestic Accident Scenarios among Ebonyi State Indigenes. IOSR Journal of Computer Engineering, 28(5), 48-58. https://doi.org/10.9790/0661-2805034858
Vancouver
Jeremiah, Chukwu, Nneka AJ, Chukwuma N. Development and Evaluation of An Explainable AI-Based Chatbot for First-Aid Assistance in Domestic Accident Scenarios among Ebonyi State Indigenes. IOSR Journal of Computer Engineering. 2026;28(5):48-58. doi:10.9790/0661-2805034858
BibTeX
@article{Jeremiah2026,
author = {Jeremiah and Chukwu and Nneka, Agwu Joy and Chukwuma, Nweke},
title = {Development and Evaluation of An Explainable AI-Based Chatbot for First-Aid Assistance in Domestic Accident Scenarios among Ebonyi State Indigenes},
journal = {IOSR Journal of Computer Engineering},
year = {2026},
volume = {28},
number = {5},
pages = {48--58},
doi = {10.9790/0661-2805034858}
}
EUVIMED