A framework detects speaker drift in TTS outputs by computing cosine similarities across speech segments and using LLMs for binary classification, supported by a human-validated synthetic benchmark.
The dawn of quantum natural language processing,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Novel Automatic Framework for Speaker Drift Detection in Synthesized Speech
A framework detects speaker drift in TTS outputs by computing cosine similarities across speech segments and using LLMs for binary classification, supported by a human-validated synthetic benchmark.