Naturalness

About: Naturalness is a research topic. Over the lifetime, 1305 publications have been published within this topic receiving 31737 citations.

Polyphone Disambiguation and Accent Prediction Using Pre-Trained Language Models in Japanese TTS Front-End

Abstract: Although end-to-end text-to-speech (TTS) models can generate natural speech, challenges still remain when it comes to estimating sentence-level phonetic and prosodic information from raw text in Japanese TTS systems. In this paper, we propose a method for polyphone disambiguation (PD) and accent prediction (AP). The proposed method incorporates explicit features extracted from morphological analysis and implicit features extracted from pre-trained language models (PLMs). We use BERT and Flair embeddings as implicit features and examine how to combine them with explicit features. Our objective evaluation results showed that the proposed method improved the accuracy by 5.7 points in PD and 6.0 points in AP. Moreover, the perceptual listening test results confirmed that a TTS system employing our proposed model as a front-end achieved a mean opinion score close to that of synthesized speech with ground-truth pronunciation and accent in terms of naturalness.

A Natural Introduction to Fine-Tuning

Abstract: A well-known topic within the philosophy of physics is the problem of fine-tuning: the fact that the universal constants seem to take non-arbitrary values in order for live to thrive in our Universe. In this paper we will talk about this problem in general, giving some examples from physics. We will review some solutions like the design argument, logical probability, cosmological natural selection, etc. Moreover, we will also discuss why it's dangerous to uphold the Principle of Naturalness as a scientific principle. After going through this paper, the reader should have a general idea what this problem exactly entails whenever it is mentioned in other sources and we recommend the reader to think critically about these concepts.

Measurement of Naturalness: Physics and Perception

Abstract: This paper presents the outcomes of an EU project (‘Measurement of Naturalness’, or MONAT), which aimed to develop a measurement system to predict the perception of naturalness for a range of materials, based on an understanding of the relationship between the physical attributes of the material and the human perceptual responses. The project included the establishment of novel measurement instrumentation to characterise the perceptually-relevant properties of the materials. New data analysis and modelling techniques were also developed to isolate the key physical properties of a material that influence the perception of naturalness, and to build mathematical models to describe the perceived naturalness in terms of these properties. The predictive power of the models proved to be extremely good, with correlation values in excess of 0.8 for new samples with properties similar to those used to develop the models, and reduction of error values above 70% for samples with properties falling outside this gamut.