What is best for spoken language understanding: small but task-dependant embeddings or huge but out-of-domain embeddings?

Sahar Ghannay, Antoine Neuraz, Sophie Rosset

Browse the full ICASSP paper archive.