Martti Vainio

Sofoklis Kakouros, Juraj Šimko, Martti Vainio et al.

This paper investigates the use of word surprisal, a measure of the predictability of a word in a given context, as a feature to aid speech synthesis prosody. We explore how word surprisal extracted from large language models (LLMs) correlates with word prominence, a signal-based measure of the salience of a word in a given discourse. We also examine how context length and LLM size affect the results, and how a speech synthesizer conditioned with surprisal values compares with a baseline system. To evaluate these factors, we conducted experiments using a large corpus of English text and LLMs of varying sizes. Our results show that word surprisal and word prominence are moderately correlated, suggesting that they capture related but distinct aspects of language use. We find that length of context and size of the LLM impact the correlations, but not in the direction anticipated, with longer contexts and larger LLMs generally underpredicting prominent words in a nearly linear manner. We demonstrate that, in line with these findings, a speech synthesizer conditioned with surprisal values provides a minimal improvement over the baseline with the results suggesting a limited effect of using surprisal values for eliciting appropriate prominence patterns.

Antti Suni, Sofoklis Kakouros, Martti Vainio et al.

Recent advances in deep learning methods have elevated synthetic speech quality to human level, and the field is now moving towards addressing prosodic variation in synthetic speech.Despite successes in this effort, the state-of-the-art systems fall short of faithfully reproducing local prosodic events that give rise to, e.g., word-level emphasis and phrasal structure. This type of prosodic variation often reflects long-distance semantic relationships that are not accessible for end-to-end systems with a single sentence as their synthesis domain. One of the possible solutions might be conditioning the synthesized speech by explicit prosodic labels, potentially generated using longer portions of text. In this work we evaluate whether augmenting the textual input with such prosodic labels capturing word-level prominence and phrasal boundary strength can result in more accurate realization of sentence prosody. We use an automatic wavelet-based technique to extract such labels from speech material, and use them as an input to a tacotron-like synthesis system alongside textual information. The results of objective evaluation of synthesized speech show that using the prosodic labels significantly improves the output in terms of faithfulness of f0 and energy contours, in comparison with state-of-the-art implementations.

Aarne Talman, Antti Suni, Hande Celikkanat et al.

In this paper we introduce a new natural language processing dataset and benchmark for predicting prosodic prominence from written text. To our knowledge this will be the largest publicly available dataset with prosodic labels. We describe the dataset construction and the resulting benchmark dataset in detail and train a number of different models ranging from feature-based classifiers to neural network systems for the prediction of discretized prosodic prominence. We show that pre-trained contextualized word representations from BERT outperform the other models even with less than 10% of the training data. Finally we discuss the dataset in light of the results and point to future research and plans for further improving both the dataset and methods of predicting prosodic prominence from text. The dataset and the code for the models are publicly available.

Antti Suni, Daniel Aalto, Martti Vainio

Prominences and boundaries are the essential constituents of prosodic structure in speech. They provide for means to chunk the speech stream into linguistically relevant units by providing them with relative saliences and demarcating them within coherent utterance structures. Prominences and boundaries have both been widely used in both basic research on prosody as well as in text-to-speech synthesis. However, there are no representation schemes that would provide for both estimating and modelling them in a unified fashion. Here we present an unsupervised unified account for estimating and representing prosodic prominences and boundaries using a scale-space analysis based on continuous wavelet transform. The methods are evaluated and compared to earlier work using the Boston University Radio News corpus. The results show that the proposed method is comparable with the best published supervised annotation methods.

Daniel Aalto, Antti Huhtala, Atle Kivelä et al.

We compare numerically computed resonances of the human vocal tract with formants that have been extracted from speech during vowel pronunciation. The geometry of the vocal tract has been obtained by MRI from a male subject, and the corresponding speech has been recorded simultaneously. The resonances are computed by solving the Helmholtz partial differential equation with the Finite Element Method (FEM). Despite a rudimentary exterior space acoustics model, i.e., the Dirichlet boundary condition at the mouth opening, the computed resonance structure differs from the measured formant structure by $\approx$ 0.7 semitones for [i] and [u] having small mouth opening area, and by $\approx$ 3 semitones for vowels [a] and [ae] that have a larger mouth opening. The contribution of the possibly open velar port has not been taken into considaration at all which adds the discrepancy for [a] in the present data set. We conclude that by improving the exterior space model and properly treating the velar port opening, it is possible to computationally attain four lowest vowel formants with an error less than a semitone. The corresponding wave equation model on MRI-produced vocal tract geometries is expected to have a comparable accuracy.