pabloaaf/Factor-TranscriptionCaseStudy: v1.0.0
| Main Author: | Pablo Angel Alvarez Fernandez |
|---|---|
| Format: | info software eJournal |
| Terbitan: |
, 2020
|
| Subjects: | |
| Online Access: |
https://zenodo.org/record/3893988 |
Daftar Isi:
- Proof of concept for further testing and development transcript generation for online classes. ABSTRACT By automating class transcriptions, we take the next step to enhance education accessibility for both presential and online courses. The computing barriers for applying speech recognition algorithms has decreased, as it is stated by the appearance of both cloud and on-premises services. It will be shown that the economics of Speech-to-Text services are no longer a barrier to their widespread usage. We provide a guide to compare those techniques and their initial, out of the box accuracy while integrating these libraries into a single test application. Our Cloud model employs AWS S3 using AWS Transcribe while the on-premises Opensource model relies on Mozilla’s DeepSpeech and an opensource built gRPC microservice based application. Moreover, beyond comparisons of Speech-to-Text libraries we also investigate the artifacts produced by high quality Speech-to-Text and what new features previously unavailable to the general student body now became accessible such as subtitles, transcripts, timestamps, and custom search features now enabled by having high quality transcripts available. In the construction of this project in order to easy deployment for testing, decisions based on using OS Containers for the greatest abstraction and portability possible were taken into consideration to tackle the non-trivial nature of application deployment. This required research into the latest web development technologies to accomplish with emphasis on the security in order to produce a reliable and secure development process and to provide open access to this proof of concept for further testing and development.