English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Automatic Speech Recognition (ASR) systems exhibit the best performance onspeech that is similar to that on which it was trained. As such,underrepresented varieties including regional dialects, minority-speakers, andlow-resource languages, see much higher word error rates (WERs) than thosevarieties seen as 'prestigious', 'mainstream', or 'standard'. This can act as abarrier to incorporating ASR technology into the annotation process forlarge-scale linguistic research since the manual correction of the erroneousautomated transcripts can be just as time and resource consuming as manualtranscriptions. A deeper understanding of the behaviour of an ASR system isthus beneficial from a speech technology standpoint, in terms of improving ASRaccuracy, and from an annotation standpoint, where knowing the likely errorsmade by an ASR system can aid in this manual correction. This work demonstratesa method of probing an ASR system to discover how it handles phonetic variationacross a number of L2 Englishes. Specifically, how particular phoneticrealisations which were rare or absent in the system's training data can leadto phoneme level misrecognitions and contribute to higher WERs. It isdemonstrated that the behaviour of the ASR is systematic and consistent acrossspeakers with similar spoken varieties (in this case the same L1) and phonemesubstitution errors are typically in agreement with human annotators. Byidentifying problematic productions specific weaknesses can be addressed bysourcing such realisations for training and fine-tuning thus making the systemmore robust to pronunciation variation.

Investigating the Sensitivity of Automatic Speech Recognition Systems to Phonetic Variation in L2 Englishes