INVESTIGADORES
EVIN Diego Alexis
congresos y reuniones científicas
Título:
N-Best Rescoring Based on Intonation Prediction for a Spanish ASR System
Autor/es:
D. EVIN, J. GURLEKIAN, H. TORRES
Lugar:
Berlín, Alemania
Reunión:
Conferencia; 21 Konferenz Elektronische Sprachsignalverarbeitung; 2010
Institución organizadora:
Beuth Hochschule für Technik Berlin
Resumen:
This paper presents a novel method for rescoring the n-best recognitionhypotheses using intonation knowledge. The model synthesizes the f0 contours foreach of the n-best hypotheses and estimates an intonative matching index betweenthe synthetic shapes and the real f0 contour. This index is applied in the rescoringprocess, and can be viewed as a degree of intonation compatibility between thehypotheses and the input sentence. The f0 prediction is based on classification andregression trees and the Fujisaki model. We evaluate our approach using a singlespeaker of the Buenos Aires Spanish LIS-SECYT database under clean and babble noisyconditions. Considering the systems under no grammar condition, the proposedmodel reduces the mean absolute word error rate in 3.1% with respect to the baselinesystem, in a consistent manner and under different noise conditions.