Project JARVIS, is a Voice-Recognition Artificial Intelligence application designed by SIS students to provide students with formative feedback on their presentations to supplement instructors’ feedback.
This application supports personalised learning by transcribing students’ presentations into text using a speech-to-text Application Programming Interface (API). The generated text and associated timestamp is subsequently analysed to generate meaningful feedback, for example, high frequency filler words, and pace of presentation compared to SMU norm.
This application also provides an interface for instructors to enable monitoring and additional feedback to students.
Project JARVIS is currently piloted in communication courses to provide personalised feedback to students on various aspects of their presentations.
In collaboration with the School of Information Systems, this was taken up by a group of students as part of the IS480 module. Utilising Speech-To-Text transcription and Natural Language Processing techniques, they developed an application capable of providing feedback on many aspects of public speaking. These include the use of filler words and their frequency of use, pace of speech relative to peers in SMU, grammatical errors, and the use of inappropriate language.
In the process, the team built a customised speech-to-text model that more accurately transcribes words specific to the Singaporean context (e.g. place and food names). This model made demonstrable improvements over commercially available and open-source Speech-To-Text transcription solutions.