Project chat-bot

Abstract

This project aims to improve on the current deployment process for Project Chat Bot’s Bani model. This is achieved by automating the building of the Question Answering agent, Bani Bot, and deployed onto the Speech Lab Chatbot Website. The current process of updating the model for deployment involves several manual maneuvers which may make maintaining the chatbot components tedious since deployment needs to be initiated whenever there is a change in dataset or changes to model configurations. Improvement to the deployment process is achieved by generalizing code structure in its components. This removes the need to amend code directly with changes to the topics and models during training and deploying the bot. Training is executed using docker, this provides the ease of training on a different machine and enable the use of docker’s volume for easy file transfer between the training and the bot. Aside from that, improvements are made to the existing chatbot website that hosts the models by introducing a model API to manage the list of listed models. Model API was introduced to better manage offered models on the website and reduce the need to amend the front-end code with the additional model. For the context of this project, question and answer pairs from the data sets, Covid-19, Baby Bonus, and Adoption Frequently Asked Question (FAQ) are used. These data sets are used to test the proposed pipeline for the building and deploying of Bani Bot onto the Speech Lab website.