Suprasegmental modelling
The aim of the paper is to show how prosodic information can be computed and used in a speech understanding system. This is illustrated within the VERBMOBIL speech-to-speech translation system, the world wide first complete system which successfully uses prosodic information in the linguistic analysis. The computed prosodic information is divided by the VERBMOBIL architecture into the following steps: feature extraction, description of the classes to be recognized, classification into these classes, and improvement of the classification results with stochastic language models. Finally, these prosodic classes are calculated in a word hypotheses graph rather than in the spoken word sequence. Prosodic information may be used at different linguistic levels, but the authors concentrate on the syntactic analysis, presenting results of extensive experiments. Prosody is used by computing probabilities for clause boundaries, accentuation, and different types of sentence mood for each of the word hypotheses computed by the word recognizer. These probabilities guide the search of the linguistic analysis. So far in VERBMOBIL, the most useful information is provided by clause boundaries, detected with a recognition rate of 94\%. For the parsing of word hypotheses graphs, the use of clause boundaries probabilities yields a speed-up of 92\% and a 96\% reduction of the alternative, readings. Other applications, addressed within the second phase of the VERBIMOBIL project, are for example the prosodic marking of accents, the prosodic marking of emotions etc., but they represent just the beginning of integrating the prosodic information into automatic speech understanding systems. For the entire collection see [Zbl 0927.00044].
- Hierarchical structure and word strength prediction of Mandarin prosody
- A metrical model of prosody for multilingual TTS
- Prosodic phrasing: Machine and human evaluation
- Automatic ToBI prediction and alignment to speed manual labeling of prosody
- Communication and prosody: Functional aspects of prosody
- On the use of prosody in automatic dialogue understanding
- Integrated recognition of words and prosodic phrase boundaries
- scientific article; zbMATH DE number 986422 (Why is no real title available?)
- On the probabilistic modelling of the form~function articulation for prosodic phenomena
- Classifying languages based on speech rhythm
- Prosody modelling in concept-to-speech generation: methodological issues
- Recognizing time pressure and cognitive load on the basis of speech: an experimental study
- scientific article; zbMATH DE number 194619 (Why is no real title available?)
- Generating prosodic attitudes in French: Data, model and evaluation
- Can we hear the prosody of smile?
- How to find trouble in communication
- The role of pause occurrence and pause duration in the signaling of narrative structure
- scientific article; zbMATH DE number 1406023 (Why is no real title available?)
This page was built for publication: Suprasegmental modelling
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4942743)
