AfriBED evaluates gender and speaker-type bias in Bemba and Nyanja ASR
Original title: AfriBED: a Speech Recognition Bias Evaluation Dataset for Low Resource African Languages
The set holds about 9 hours of Bemba and 18 hours of Nyanja from 80 and 127 speakers.
AfriBED is a speech-recognition bias evaluation dataset for Bemba and Nyanja, two low-resource African languages. It supplies comparable speech so automatic speech recognition systems can be checked for gender and speaker-type bias. Baseline experiments indicate the data can surface those attribute biases.
- The Bemba portion has 3,928 utterances and 9 hours from 80 speakers.
- The Nyanja portion has 6,260 utterances and about 18 hours from 127 speakers.
- The target attributes are gender and speaker type.
From the original article
1 min read at arxiv.orgDespite advancements in automatic speech recognition systems (ASR), current systems exhibit speaker-attribute bias, i. e. , their speech recognition performance is biased against certain speakers based on their demographic attributes such as gender, accents, and speaker type.
By Claytone Sikasote · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.