L2-ARCTIC
A publicly available dataset designed for second-language English speech research. It contains English recordings from speakers whose native languages include Arabic, Mandarin, Hindi, Korean, Spanish, and Vietnamese, with each speaker contributing approximately 1,150 utterances. It is one of the most widely used benchmarks in accent classification research.
Timeline (last 90 days)
The study uses L2-ARCTIC and multiple external English accent datasets to compare two evaluation settings: file-level splitting and speaker-isolated splitting.
Models trained solely on external data cannot reliably generalize to the L2-ARCTIC dataset, indicating significant domain mismatch between different corpora.
L2-ARCTIC is a publicly available dataset designed for second-language English speech research, containing speakers with native language backgrounds including Arabic, Mandarin, Hindi, Korean, Spanish, and Vietnamese, each contributing approximately 1,150 utterances.
All Facts (3)
The study uses L2-ARCTIC and multiple external English accent datasets to compare two evaluation settings: file-level splitting and speaker-isolated splitting.
50%UnverifiedModels trained solely on external data cannot reliably generalize to the L2-ARCTIC dataset, indicating significant domain mismatch between different corpora.
50%UnverifiedL2-ARCTIC is a publicly available dataset designed for second-language English speech research, containing speakers with native language backgrounds including Arabic, Mandarin, Hindi, Korean, Spanish, and Vietnamese, each contributing approximately 1,150 utterances.
50%