Multilingual speech recognition
The paper presents the architecture of the SQEL (Spoken Queries in European Languages) dialog system for language identification in the context of the multi-lingual and multi-functional speech understanding task. National (four) language SQEL demonstrators are based on the EVAR database and system that can handle continuously spoken German and English inquiries on air flight and train connections. The multi-lingual SQEL demonstrators are built on two basic approaches to language identification: (1) The first approach develops an explicit module for language identification, based on training a common codebook for all the considered languages, and integrating over the output probability of language specific \(n\)-gram models trainer over the codebook sequences. This approach allows to recognize languages that the system can not process and give out a pre-recorded message in that language. (2) In the second approach, the trained recognizers of the languages to be recognized, the lexicons, and the language models are combined into one multi-lingual recognizer. Allowing only transitions between the words from one language, each hypothesized word chain contains words from just one language, and the language identification is an explicit by-product of the speech recognizer. The first experimental results on the application of the two basic approaches to language identification are reported. For the entire collection see [Zbl 0927.00044].
- Multilingual speech recognition in seven languages
- Automatic language identification
- Multilingual phone models for vocabulary-independent speech recognition tasks
- Towards multilingual interoperability in automatic speech recognition
- Language-independent and language-adaptive acoustic modeling for speech recognition
- Towards multilingual interoperability in automatic speech recognition
- Multilingual phone models for vocabulary-independent speech recognition tasks
- Multilingual speech recognition in seven languages
- Automatic language identification
- Building multilingual spoken dialogue systems
- scientific article; zbMATH DE number 1897867 (Why is no real title available?)
- A novel large vocabulary continuous speech recognition algorithm combined with language recognition
This page was built for publication: Multilingual speech recognition
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4942759)
