Models of Visually Grounded Speech Signal Pay Attention to Nouns: A Bilingual Experiment on English and Japanese
Fiche du document
12 mai 2019
- handle: 10670/1.wmbrx4
- hal: hal-02013984
- ARXIV: 1902.03052
- doi: 10.1109/ICASSP.2019.8683069
Ce document est lié à :
info:eu-repo/semantics/altIdentifier/arxiv/1902.03052
Ce document est lié à :
info:eu-repo/semantics/altIdentifier/doi/10.1109/ICASSP.2019.8683069
info:eu-repo/semantics/OpenAccess
Sujets proches
Concentration (Psychology)Citer ce document
William N Havard et al., « Models of Visually Grounded Speech Signal Pay Attention to Nouns: A Bilingual Experiment on English and Japanese », HAL-SHS : linguistique, ID : 10.1109/ICASSP.2019.8683069
Métriques
Partage / Export
Résumé
We investigate the behaviour of attention in neural models of visually grounded speech trained on two languages: English and Japanese. Experimental results show that attention focuses on nouns and this behaviour holds true for two very typologically different languages. We also draw parallels between artificial neural attention and human attention and show that neural attention focuses on word endings as it has been theorised for human attention. Finally, we investigate how two visually grounded monolingual models can be used to perform cross-lingual speech-to-speech retrieval. For both languages, the enriched bilingual (speech-image) corpora with part-of-speech tags and forced alignments are distributed to the community for reproducible research. Index Terms-grounded language learning, attention mechanism , cross-lingual speech retrieval, recurrent neural networks.