6533b7d0fe1ef96bd125a4a2
RESEARCH PRODUCT
ASR performance prediction on unseen broadcast programs using convolutional neural networks
Benjamin LecouteuxLaurent BesacierZied ElloumiOlivier GalibertJuliette Kahnsubject
FOS: Computer and information sciencesComputer Science - Computation and LanguageComputer scienceSpeech recognitionFeature extractionInformationSystems_INFORMATIONSTORAGEANDRETRIEVAL02 engineering and technology010501 environmental sciences01 natural sciencesConvolutional neural network[INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]Task (project management)[INFO.INFO-CL] Computer Science [cs]/Computation and Language [cs.CL]0202 electrical engineering electronic engineering information engineeringTask analysisPerformance prediction020201 artificial intelligence & image processingMel-frequency cepstrumTranscription (software)Hidden Markov modelComputation and Language (cs.CL)ComputingMilieux_MISCELLANEOUS0105 earth and related environmental sciencesdescription
In this paper, we address a relatively new task: prediction of ASR performance on unseen broadcast programs. We first propose an heterogenous French corpus dedicated to this task. Two prediction approaches are compared: a state-of-the-art performance prediction based on regression (engineered features) and a new strategy based on convolutional neural networks (learnt features). We particularly focus on the combination of both textual (ASR transcription) and signal inputs. While the joint use of textual and signal features did not work for the regression baseline, the combination of inputs for CNNs leads to the best WER prediction performance. We also show that our CNN prediction remarkably predicts the WER distribution on a collection of speech recordings.
| year | journal | country | edition | language |
|---|---|---|---|---|
| 2018-04-23 |