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/

Zendy Plus

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

Zendy Tools

Zendy Open

This paper presents recent progress in the acoustic modelling ofunder-resourced code-switched (CS) speech in multiple South African languages.We consider two approaches. The first constructs separate bilingual acousticmodels corresponding to language pairs (English-isiZulu, English-isiXhosa,English-Setswana and English-Sesotho). The second constructs a single unifiedfive-lingual acoustic model representing all the languages (English, isiZulu,isiXhosa, Setswana and Sesotho). For these two approaches we consider theeffectiveness of semi-supervised training to increase the size of the verysparse acoustic training sets. Using approximately 11 hours of untranscribedspeech, we show that both approaches benefit from semi-supervised training. Thebilingual TDNN-F acoustic models also benefit from the addition of CNN layers(CNN-TDNN-F), while the five-lingual system does not show any significantimprovement. Furthermore, because English is common to all language pairs inour data, it dominates when training a unified language model, leading toimproved English ASR performance at the expense of the other languages.Nevertheless, the five-lingual model offers flexibility because it can processmore than two languages simultaneously, and is therefore an attractive optionas an automatic transcription system in a semi-supervised training pipeline.

Semi-supervised acoustic model training for five-lingual code-switched ASR