A challenge system combining speaker diarization, speaker embeddings, and a Qwen2.5 LLM adapter architecture reports 18.08% tcpWER on multilingual multi-speaker ASR, far below the 60.39% baseline.
All systems are evaluated on the development set, while only average scores are available on the test set via the official evaluation server
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
other 1
citation-polarity summary
fields
eess.AS 1years
2025 1verdicts
CONDITIONAL 1roles
other 1polarities
unclear 1representative citing papers
citing papers explorer
-
The DKU System for Multi-Speaker Automatic Speech Recognition in MLC-SLM Challenge
A challenge system combining speaker diarization, speaker embeddings, and a Qwen2.5 LLM adapter architecture reports 18.08% tcpWER on multilingual multi-speaker ASR, far below the 60.39% baseline.