Statistical Machine Translation from Arab Vocal Improvisation to Instrumental Melodic Accompaniment

Fadi Al-Ghawanmeh 1 Kamel Smaïli 2
2 SMarT - Statistical Machine Translation and Speech Modelization and Text
LORIA - NLPKD - Department of Natural Language Processing & Knowledge Discovery
Abstract : Vocal improvisation is an essential practice in Arab music. The interactivity between the singer and the instru-mentalist(s) is a main feature of this deep-rooted musical form. As part of the interactivity, the instrumentalist re-capitulates, or translates, each vocal sentence upon its completion. In this paper, we present our own parallel corpus of instrumentally accompanied Arab vocal improvisation. The initial size of the corpus is 2779 parallel sentences. We discuss the process of building this corpus as well as the choice of data representation. We also present some statistics about the corpus. Then we present initial experiments on applying statistical machine translation to propose an automatic instrumental accompaniment to Arab vocal improvisation. The results with this small corpus, in comparison to classical machine translation of natural languages, are very promising: a BLEU of 24.62 from Vocal to instrumental and 24.07 from instrumental to vocal.
Type de document :
Communication dans un congrès
ICNLSSP 2017 - International Conference on Natural Language, Signal and Speech Processing, Dec 2017, Casablanca, Morocco. 2017, 〈http://icnlssp.isga.ma〉
Liste complète des métadonnées

Littérature citée [15 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/hal-01660023
Contributeur : Kamel Smaïli <>
Soumis le : samedi 9 décembre 2017 - 18:04:28
Dernière modification le : mardi 24 avril 2018 - 13:29:50

Fichier

ICNLSSP2017_paper_27.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01660023, version 1

Citation

Fadi Al-Ghawanmeh, Kamel Smaïli. Statistical Machine Translation from Arab Vocal Improvisation to Instrumental Melodic Accompaniment. ICNLSSP 2017 - International Conference on Natural Language, Signal and Speech Processing, Dec 2017, Casablanca, Morocco. 2017, 〈http://icnlssp.isga.ma〉. 〈hal-01660023〉

Partager

Métriques

Consultations de la notice

111

Téléchargements de fichiers

35