Public Speaking Web Trainer

This paper describes a web-based application that can be used to prepare for giving a public talk. Firstly, an overview of existing public speaking applications is provided. We noticed that existing applications lack evaluation customization, presentation file handling features, and integration with external services such as learning management systems. This application provides feedback based on a pack of criteria such as speech duration, speech pace, filler words usage, and others. We propose flexible application architecture so new steps of evaluation can be added, removed, or changed, also it can be scaled to support potentially increasing application load. We measured training processing time both for a single training and a set of trainings submitted to the application simultaneously. So far the longest step of processing is speech recognition so configurations with different number of speech recognition instances were considered to get faster training processing time.