Modeling Improved Prosody Generation from High-Level Linguistically Annotated Corpora

Abstract

SUMMARY Synthetic speech usually suffers from bad F0 contour surface. The prediction of the underlying pitch targets robustly relies on the quality of the predicted prosodic structures, i.e. the corresponding sequences of tones and breaks. In the present work, we have utilized a linguistically enriched annotated corpus to build data-driven models for predicting prosodic structures with increased accuracy. We have then used a linear regression approach for the F0 modeling. An appropriate XML annotation scheme has been introduced to encode syntax, grammar, new or already given information, phrase subject/object information, as well as rhetorical elements in the corpus, by exploiting a Natural Language Generator (NLG) system. To prove the benefits from the introduction of the enriched input meta-information, we first show that while tone and break CART predictors have high accuracy when standing alone (92.35% for breaks, 87.76% for accents and 99.03% for endtones), their application in the TtS chain degrades the Linear Regression pitch target model. On the other hand, the enriched linguistic meta-information minimizes errors of models leading to a more natural F0 surface. Both objective and subjective evaluation were adopted for the intonation contours by taking into account the propagated errors introduced by each model in the synthesis chain. key words : prosody modeling, text-to-speech, linguistic meta-information, synthetic prosody evaluation

Experimental Evaluation of Tree-Based Algorithms for Intonational Breaks Representation

“…Further improvements in the tree-structured predictors can be achieved by the introduction of more delicate linguistic features as has been inspected on [16] for the CART approach.…”

Differences Among Sighted Individuals and Individuals with Visual Impairments in Word Intelligibility Presented via Synthetic and Natural Speech

“…The audio recordings were digitized through the audio card. DEMOSTHe´NES is a modular and scalable (Xydas & Kouroupetroglou, 2001a), multilingual and polyglot (Xydas & Kouroupetroglou, 2001b) TTS system that supports Greek and English with various voices and incorporates advanced speech synthesis methodologies in order to produce almost natural pitch and prosody (Xydas, Spiliotopoulos, & Kouroupetroglou, 2005;Xydas, Zervas, Kouroupetroglou, Fakotakis, & Kokkinakis, 2005). The DEMOSTHe´NES TTS system used in this study incorporates a diphone-based MBROLA (Multi-Band Resynthesis OverLap and Process) synthesis technique (Dutoit, Pagel, Pierret, Bataille, & van Der Vreken, 1996), with the female voice and the following parameters: pitch ¼ 210Hz; speed ¼ 95 wpm.…”

Prosodic Boundary Prediction for Greek Speech Synthesis

“…The labelling of the intonational phenomena had been conducted main ly by listening to the recorded utterance in conjunction to observation of amplitude and pitch contour of the speech signal. The annotator"s transcription consistency was further evaluated by cross checking statistically our data with a prosodic corpus constructed at the University of Athens for speech synthesis purposes [32].…”