Correcting automatic speech recognition captioning errors in real time

Abstract

Lectures can be digitally recorded and replayed to provide multimedia revision material for students who attended the class and a substitute learning experience for students unable to attend. Deaf and hard of hearing people can find it difficult to follow speech through hearing alone or to take notes while they are lip-reading or watching a sign-language interpreter. Synchronising the speech with text captions can ensure deaf students are not disadvantaged and assist all learners to search for relevant specific parts of the multimedia recording by means of the synchronised text. Automatic speech recognition has been used to provide real-time captioning directly from lecturers' speech in classrooms but it has proved difficult to obtain accuracy comparable to stenography. This paper describes the development, testing and evaluation of a system that enables editors to correct errors in the captions as they are created by automatic speech recognition and makes suggestions for future possible improvements.

Improving Speech Recognition with the Robot Interaction Language

“…31,32 This value might change slightly for different languages and speech recognizers. Ideally, we would have liked to confirm this value in the context of ROILA dictation tasks, but due to time constraints, we had to accept this estimation as is.…”

Third-party error detection support mechanisms for dictation speech recognition

“…Thirdparty error detection, which is the focus of the current research, can be particularly challenging. For example, (Wald et al, 2007) reported that a third-party human editor was able to correct a fraction of the errors in speech recognition output in real time.…”

“…Detecting and selecting errors consume 75% of participants' time in error correction (Wald et al, 2007). Thus, it is desirable to reduce the cognitive workload associated with error detection.…”

The design of human-powered access technology

“…For example, OCR seems like a solved problem until it fails to decipher the text on a road sign captured by a cell phone camera [33], object recognition works reasonably well until the camera is held by a blind person [15,28,38], and the laudable 99% accuracy reported by commercial automatic speech recognition systems [4] falls off precipitously on casual conversation or any time it has not been trained for the speaker [45]. Even the automatic techniques used by the screenreading software to convey the contents of the computer screen to blind people are error-prone, unreliable, and, therefore, confusing [16,17,30].…”