| dc.description.abstract |
A chatbot is a software tool used to interact with customers using text or speech in a natural language processing environment. In Ethiopia, one of the major concerns is providing high-quality, affordable healthcare services to its growing population. However, challenges such as transportation shortages, lack of medical professionals, and limited infrastructure hinder the delivery of healthcare services, leading to increased mortality rates and delayed treatments. The use of chatbots in various industries, including healthcare, has been rapidly growing in recent years. Thus, creating a bilingual medical chatbot capable of diagnosing, providing follow-ups, and offering advice for patients with Diabetes Mellitus using Deep Learning Techniques without the need for a doctor's visit is crucial. This chatbot aims to assist individuals proficient in English or Afaan Oromo languages, with a focus on rural and underserved communities.
Deep learning refers to the process by which neural networks analyze vast datasets, particularly those with multiple layers. AI-powered chatbots can serve as automated conversational agents to advance healthcare. For this study, a dataset was created by compiling frequently asked questions from websites into JSON files in both English and Afaan Oromo languages. Sequential deep neural network chatbot models were developed using Natural Language Processing techniques. Model performance was evaluated using accuracy metrics, with the initial model achieving 87.96% accuracy before applying regularization techniques. To address model overfitting, L2 regularization was implemented, resulting in an accuracy score of 92.59% after regularization. Human evaluation was conducted in addition to metric evaluation, assessing factors such as the number of queries generated by the model, the number of responses provided by the system, and user acceptance. Criteria including attractiveness, response time, user-friendliness, efficiency, and system viability were used to evaluate the system's performance. |
en_US |
