eduzhai > Applied Sciences > Engineering >

Gender in Danger? Evaluating Speech Translation Technology on the MuST-SHE Corpus

  • Save

... pages left unread,continue reading

Document pages: 12 pages

Abstract: Translating from languages without productive grammatical gender like Englishinto gender-marked languages is a well-known difficulty for machines. Thisdifficulty is also due to the fact that the training data on which models arebuilt typically reflect the asymmetries of natural languages, gender biasincluded. Exclusively fed with textual data, machine translation isintrinsically constrained by the fact that the input sentence does not alwayscontain clues about the gender identity of the referred human entities. Butwhat happens with speech translation, where the input is an audio signal? Canaudio provide additional information to reduce gender bias? We present thefirst thorough investigation of gender bias in speech translation, contributingwith: i) the release of a benchmark useful for future studies, and ii) thecomparison of different technologies (cascade and end-to-end) on two languagedirections (English-Italian French).

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...