Talko is a web solution that provides instant feedback on your public speaking skills and habits. All you have to do is input your speech transcript and start talking! Talko will analyze your energy levels, enunciation, pace, pause frequency, and more to provide you with detailed and constructive feedback.
For the backend, we built a centralized RESTful Flask API to fetch all backend data from one endpoint. We used Google Cloud Storage to store files greater than 30 seconds as we found that locally saved audio files could only retain about 20-30 seconds of audio. We also used Google Cloud App Engine to deploy our Flask API as well as Google Cloud Speech To Text to transcribe the audio. Various python libraries were used for the analysis of voice data, and the resulting response returns within 5-10 seconds. The web application user interface was built using React, HTML and CSS and focused on displaying analyses in a clear and concise manner. We had two members of the team in charge of designing and developing the front end and two working on the back end functionality.
By analyzing speaking data from many successful public speakers from a variety industries and backgrounds, we have established relatively robust standards for optimal speed, energy levels and pausing frequency during a speech. Taking into consideration the overall tone of the speech, as selected by the user, we are able to tailor our analyses to the user's needs. This simple and easy to use web application will offer users insight into their overall accuracy, enunciation, WPM, pause frequency, energy levels throughout the speech, error frequency per interval and summarize some helpful tips to improve their performance the next time around.