Tag Archives: auto-captions

Publications

New article in EOS on best practices for AI-based auto-captioning

Our new opinion article in EOS presents evidence based best practices for getting good quality AI-based live auto-captions. We measured the Word Error Rate for jargon rich structural geology passages to see what conditions produced the best quality captions.

Citation: Cooke, M., C. R. Child, E. C. Sibert, C. von Hagke, and S. G. Zihms (2020), Caption this! Best practices for live captioning presentations, Eos, 101, https://doi.org/10.1029/2020EO150246. Published on 09 October 2020.

photo of a screen showing plot of caption performance. Using an external microphone and speaking deliberately produced the lowest word error rate. non-native accented english produced a range of error.
Skip to toolbar