TOBI: a standard for labeling English prosody

By Kim E. A. Silverman, Mary E. Beckman, J. Pitrelli, Mari Ostendorf, Colin W. Wightman, P. Price, Janet B. Pierrehumbert, Julia Hirschberg. 1992

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Xu Tan
2. Tao Qin
3. Frank K. Soong
4. Yu Zhang
5. Junichi Yamagishi
6. Wei-Ning Hsu
7. Xin Eric Wang
8. Tie-Yan Liu

Citations in the corpus

• A Survey on Neural Speech Synthesis (2021, methodology): “A lot of works [312, 186, 137, 277] investigate different models and features to predict the prosody tags based on ToBI.”
• Towards End-to-End Prosody Transfer for Expressive Speech Synthesis with Tacotron (2018, methodology): “Numerous works have explored annotation schemes for diagramming and automatic labeling of prosody: ToBi (Silverman et al., 1992), AuToBI (Rosenberg, 2010), Tilt (Taylor, 1998), INTSINT (Hirst, 2001), SLAM (Obin et al., 2014) all describe methods for the annotation and automatic extraction of labels ”
• Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis (2018, methodology): “AuToBI, for example, (Rosenberg, 2010) aims to produce ToBI (Silverman et al., 1992) labels that can be used by other TTS models.”
• SLAM: Automatic Stylization and Labelling of Speech Melody (2014, background): “The transcription of speech prosody aims at representing the variations of speech prosody that are considered as relevant with an alphabet of elementary symbols [1, 2, 3, 4], which each instantiates a function in the speech communication process.”
• Exploiting prosodic breaks in language modeling with random forests (2008, background): “) New opportunities of using prosody emerged after the availability of a prosodically labeled speech corpus [10], where tones and breaks were hand-labeled with a subset of ToBI labeling scheme [11].”
• Articulatory and acoustic studies on domain-initial strengthening in Korean (2001, background): “Crucial for our purposes is the idea that adjacent higher constituents show greater disjuncture than adjacent lower constituents, as coded, for example, by the Break Indices of ToBI transcription systems (e.g., Silverman, Beckman, Pitrelli, Ostendorf, Wightman, Price, Pierrehumbert & Hirschberg, 19”
• Parsing Speech: a Neural Approach to Integrating Lexical and Acoustic-Prosodic Information (2017, methodology): “the annotation standard used for Switchboard data: The root node of the tree is an “S” node although it is not a complete sentence. ers (prosodic breaks) based on the ToBI prosodic annotation system (Silverman et al., 1992). In some studies, automatically predicted prosodic breaks are incorporated d”
• Expressive TTS Training With Frame and Style Reconstruction Loss (2020, background): “Even if a labeling scheme is possible [29], [30], a set of discrete labels may not be sufficient”
• Emotional speech recognition: Resources, features, and methods (2006, background): “the ToBI (Tones and Breaks Indices) system (Silverman et al., 1992).”
• Dependency Parsing based Semantic Representation Learning with Graph Neural Network for Enhancing Expressiveness of Text-to-Speech (2021, methodology): “The input of E2E-TTS systems is usually enriched by context-free syntactic information using taggers and syntax parsers [10, 11, 12].”
• The Spoken Dutch Corpus. Overview and First Evaluation (2000, methodology): “The scheme was developed by C. Gussenhoven, T. Rietveld, and J. Terken and resembles the ToBI (Tones and Break Indices; Silverman et al .”
• The original ToBI system and the evolution of the ToBI framework (2003, methodology): “The contrast also causes more inter-transcriber disagreement in transcribing MAE than any other accent pair (see Silverman et al. 1992; Pitrelli, Beckman and Hirschberg 1994; Syrdal et al. 2001).”
• Robust and Fine-grained Prosody Control of End-to-end Speech Synthesis (2018, methodology): “Prosody modeling had been done in a supervised manner by using annotated labels, such as those in ToBI [7].”
• Experimental and theoretical advances in prosody: A review (2010, unknown): “ToBI was originally developed to transcribe the intonation of American English but has since been adapted to transcribe a wide variety of languages (Jun, 2005).”
• A Vector Quantized Variational Autoencoder (VQ-VAE) Autoregressive Neural $F_0$ Model for Statistical Parametric Speech Synthesis (2020, methodology): “It contained 40 dilated CNN blocks with the dilation size of the n-th block as 2mod(n−1,10), n ∈ [1, 40].”