Unsupervised Word Segmentation from Speech with Attention

Abstract : We present a first attempt to perform attentional word segmen-tation directly from the speech signal, with the final goal to automatically identify lexical units in a low-resource, unwritten language (UL). Our methodology assumes a pairing between recordings in the UL with translations in a well-resourced language. It uses Acoustic Unit Discovery (AUD) to convert speech into a sequence of pseudo-phones that is segmented using neural soft-alignments produced by a neural machine translation model. Evaluation uses an actual Bantu UL, Mboshi; comparisons to monolingual and bilingual baselines illustrate the potential of attentional word segmentation for language documentation.
Document type :
Conference papers
Liste complète des métadonnées

Cited literature [22 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01818092
Contributor : Laurent Besacier <>
Submitted on : Monday, June 18, 2018 - 4:41:32 PM
Last modification on : Thursday, April 4, 2019 - 10:18:05 AM
Document(s) archivé(s) le : Wednesday, September 26, 2018 - 6:25:36 PM

File

template.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01818092, version 1

Citation

Pierre Godard, Marcely Zanon Boito, Lucas Ondel, Alexandre Berard, François Yvon, et al.. Unsupervised Word Segmentation from Speech with Attention. Interspeech 2018, Sep 2018, Hyderabad, India. ⟨hal-01818092⟩

Share

Metrics

Record views

290

Files downloads

204