Gender and Dialect Bias in YouTube's Automatic Captions
نویسنده
چکیده
This project evaluates the accuracy of YouTube’s automatically-generated captions across two genders and five dialects of English. Speakers’ dialect and gender was controlled for by using videos uploaded as part of the “accent tag challenge”, where speakers explicitly identify their language background. The results show robust differences in accuracy across both gender and dialect, with lower accuracy for 1) women and 2) speakers from Scotland. This finding builds on earlier research finding that speaker’s sociolinguistic identity may negatively impact their ability to use automatic speech recognition, and demonstrates the need for sociolinguistically-stratified validation of systems.
منابع مشابه
Effects of Talker Dialect, Gender & Race on Accuracy of Bing Speech and YouTube Automatic Captions
This project compares the accuracy of two automatic speech recognition (ASR) systems–Bing Speech and YouTube’s automatic captions–across gender, race and four dialects of American English. The dialects included were chosen for their acoustic dissimilarity. Bing Speech had differences in word error rate (WER) between dialects and ethnicities, but they were not statistically reliable. YouTube’s a...
متن کاملWho's funny: gender stereotypes, humor production, and memory bias.
It has often been asserted, by both men and women, that men are funnier. We explored two possible explanations for such a view, first testing whether men, when instructed to be as funny as possible, write funnier cartoon captions than do women, and second examining whether there is a tendency to falsely remember funny things as having been produced by men. A total of 32 participants, half from ...
متن کاملA Study of Inflectional Categories of Noun in Sistani Dialect
The present article aims to provide a synchronic study of the inflectional or morpho-syntactic categories of noun in Sistani dialect. These categories comprise person, number, gender or noun class, definiteness, case, and possession. Linguistic data was collected via recording free speech, and interviewing with 30 (15 females, 15 males) illiterate Sistani language consultants of age 40–102 year...
متن کامل"Oh, I've Heard That Before": Modelling Own-Dialect Bias After Perceptual Learning by Weighting Training Data
Human listeners are able to quickly and robustly adapt to new accents and do so by using information about speaker’s identities. This paper will present experimental evidence that, even considering information about speaker’s identities, listeners retain a strong bias towards the acoustics of their own dialect after dialect learning. Participants’ behaviour was accurately mimicked by a classifi...
متن کاملBilinguality vs. Monolinguality among Kalhuri Kurdish Speakers: Gender, Social Class and English Language Achievement
Today in multilingual contexts, many parents prefer to rear their children in the dominant language rather than in their mother tongue. This phenomenon is widespread among native speakers of Kalhuri dialect of the Kurdish language in the multilingual context of Iran, too. Nevertheless, some studies have evidenced the privilege of bilinguals in learning an additional language though some others ...
متن کامل