Optimization of Feminacare Chatbot Application Using SeaLLM Model

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Feminacare, a women's health consultation application with a chatbot, previously used the LSTM model, which had an accuracy of 61% but often gave less relevant responses. This research proposes the use of the SeaLLM model with the Retrieval-Augmented Generation (RAG) approach to improve the accuracy of the chatbot. Based on three trials in the evaluation of 120 questions, the chatbot obtained a mean accuracy of 87%, with a precision of 93%, recall of 89%, and F1-score of 91%. Compared to previous models, this approach produced more relevant and accurate responses. Overall, this research proves that the application of SeaLLM with RAG can improve the effectiveness of chatbots in providing women's health information. However, further improvements are still needed, especially in handling more complex questions by expanding the dataset.

This publication has 15 references indexed in Scilit: