Language technology practitioners as language managers: arbitrating data bias and predictive bias in ASR

Markl, Nina; McNulty, Stephen Joseph

doi:10.48550/arxiv.2202.12603

Cited by 2 publications

(2 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In automatic speech recognition, biases are in the direction of better recognition of 'standard' accents Harrington, 2023), one or other of male or female voices depending on the system (Markl and McNulty, 2022) as well as non-pathological voices (Benzeghiba et al, 2007;Markl and McNulty, 2022). Additionally, as noted by Benzeghiba et al (2007), children's voices and elderly voices are also generally not modelled well and cause performance issues with ASR.…”

Section: Fundingmentioning

confidence: 99%

Capturing Talk: The Institutional Practices Surrounding the Transcription of Spoken Language

2024

Frontiers Research Topics

View full text Add to dashboard Cite

Frontiers is more than just an open access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals. Frontiers journal seriesThe Frontiers journal series is a multi-tier and interdisciplinary set of openaccess, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers journal series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too. Dedication to qualityEach Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public -and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews. Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation. What are Frontiers Research Topics?Frontiers Research Topics are very popular trademarks of the Frontiers journals series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area.

show abstract

Section: Fundingmentioning

confidence: 99%

Capturing Talk: The Institutional Practices Surrounding the Transcription of Spoken Language

2024

Frontiers Research Topics

View full text Add to dashboard Cite

show abstract

“…In automatic speech recognition, biases are in the direction of better recognition of 'standard' accents (Markl, 2022;Wassink et al, 2022;Harrington, 2023), one or other of male or female voices depending on the system (Markl and McNulty, 2022) as well as non-pathological voices (Benzeghiba et al, 2007;Markl and McNulty, 2022). Additionally, as noted by Benzeghiba et al (2007), children's voices and elderly voices are also generally not modelled well and cause performance issues with ASR.…”

Section: Introductionmentioning

confidence: 99%

Automatic speech recognition and the transcription of indistinct forensic audio: how do the new generation of systems fare?

Loakes

2024

Front. Commun.

View full text Add to dashboard Cite

This study provides an update on an earlier study in the “Capturing Talk” research topic, which aimed to demonstrate how automatic speech recognition (ASR) systems work with indistinct forensic-like audio, in comparison with good-quality audio. Since that time, there has been rapid technological advancement, with newer systems having access to extremely large language models and having their performance proclaimed as being human-like in accuracy. This study compares various ASR systems, including OpenAI’s Whisper, to continue to test how well automatic speaker recognition works with forensic-like audio. The results show that the transcription of a good-quality audio file is at ceiling for some systems, with no errors. For the poor-quality (forensic-like) audio, Whisper was the best performing system but had only 50% of the entire speech material correct. The results for the poor-quality audio were also generally variable across the systems, with differences depending on whether a .wav or .mp3 file was used and differences between earlier and later versions of the same system. Additionally, and against expectations, Whisper showed a drop in performance over a 2-month period. While more material was transcribed in the later attempt, more was also incorrect. This study concludes that forensic-like audio is not suitable for automatic analysis.

show abstract

Language technology practitioners as language managers: arbitrating data bias and predictive bias in ASR

Cited by 2 publications

References 37 publications

Capturing Talk: The Institutional Practices Surrounding the Transcription of Spoken Language

Capturing Talk: The Institutional Practices Surrounding the Transcription of Spoken Language

Automatic speech recognition and the transcription of indistinct forensic audio: how do the new generation of systems fare?

Contact Info

Product

Resources

About