Detection of Input-Difficult Words by Automatic Speech Recognition for PC Captioning

Takeuchi, Yoshinori; Kojima, Daiki; Sano, Sachiko; Kanamori, Shinji

doi:10.1007/978-3-319-94277-3_32

Cited by 1 publication

(1 citation statement)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…12 A human captioner could also ask for things known to be more difficult to catch in a lecture and to be provided with these ahead of time, for example, technical terms, proper nouns, and acronyms or other terms that contain letters or numbers. 15 Human captioners can be very accurate if they caption after production, where they are watching a saved video and thus can slow it down or speed it up as needed and repeat sections where the speech is unclear. Human captioners also ensure that the spelling, grammar, and punctuation are accurately transcribed so as to create an easily readable document.…”

Section: Introductionmentioning

confidence: 99%

Comparison of accuracy of machine-generated or human-generated captions of Zoom live lectures in a comparative theriogenology course

Kustritz

Rupprecht

Zhitnitskiy

2023

View full text Add to dashboard Cite

Captions are available with captured lectures for student review. In this study, automatically generated captions from Zoom, Kaltura, and YouTube were compared for accuracy with captions generated by a human being. Also investigated was the effect of speaker on accuracy of captioning – does the speed with which someone speaks or their accent alter accuracy of captioning? YouTube was by far the most accurate of the automatic captioning systems. There were numerous mistakes made by Zoom and Kaltura and some significantly altered meaning. Mistakes were due to the transcribing systems, not to things specific to the presenters. Instructors should get in the habit of reviewing tran scripts of their lectures to ensure students are not misled.

show abstract

Section: Introductionmentioning

confidence: 99%