End-to-end ASR for code-switched Hindi-English with <50 hours of data shows gains from multi-task learning and corpus balancing but underperforms cascaded baselines.
Data Table 1 describes the conversational Hindi-English data used in our study; more details can be found in [23]
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
End-to-End ASR for Code-switched Hindi-English Speech
End-to-end ASR for code-switched Hindi-English with <50 hours of data shows gains from multi-task learning and corpus balancing but underperforms cascaded baselines.