A Voice-Assisted Damage Reporting System
The traditional method of reporting car damages in car companies involves using report sheets that are filled manually by a specialist then transformed into digital formats to be stored. This process could be very costly in terms of money and time especially in big companies. In order to reduce this cost, we propose a solution that uses direct human’s voice to report car damages. A Machine Learning text classifier is used in combination with a Speech-to-Text service to build and store an accurate damage report automatically. We mainly focused on the text classifier component of the system where we experimented with multiple ML classification algorithms and performed additional model tuning on the most promising one. The classifiers were trained and tested using a dataset that was built specifically for the purpose of this project. The final version of the system consisted of a Neural Networks model and produced a 99.78% classification accuracy, which is considered very good despite some limitations that are discussed throughout this dissertation.