Semantic similarity for detecting recognition errors in automatic speech transcripts
HLT '05 Proceedings of the conference on Human Language Technology and Empirical Methods in Natural Language Processing
Combination of error detection techniques in automatic speech transcription
AIS'11 Proceedings of the Second international conference on Autonomous and intelligent systems
Hi-index | 0.00 |
Recognizer Output Voting Error Reduction (ROVER), is a well-known procedure for decoders' output combination aiming at reducing the Word Error Rate (WER) in large vocabulary transcriptions. This paper presents a novel voting scheme to boost the current ROVER's performance. A contextual analysis is carried out to compute robust confidence scores, which are used during the voting stage. The confidence scores are collected from several automatic error detectors. Experiments have proven that it is possible to outperform the ROVER voting schemes.