Language Support

NSC's Language Approach – The Language Packages

NSC's speech recognition engines perform phoneme-based, speaker-independent recognition allowing for easy adoption of the technology since training is done once per language at NSC. The speech engines themselves are language independent. The language-specific information is in a language package including language-specific linguistic information, acoustic models, and lexicons.

NSC's products currently support the following languages:

  • English US
  • English UK
  • Spanish US
  • Spanish Columbia
  • Hebrew
  • Arabic (various dialects)
  • Farsi
  • Russian
  • French (Moroccan)
  • German
  • Italian
  • Turkish
  • Greek
  • Portuguese (Brazil)
  • Korean

As part of NSC's roadmap, the company is continuously adding support for additional languages. Through participation in several major international consortia with access to language resources in more than 50 languages, NSC can offer support of new languages in a short period of time. These consortia use the SpeechDat standards and specifications for data collection. To date, NSC has participated in the following consortia for language data collection:

  • Orientel - Language resources for the Mediterranean and Middle East regions.
  • SALA - SpeechDat across Latin America
  • LILA - A collection of language resources for languages in the Asian Pacific region
  • SPEECON - Speech-driven interfaces for consumer devices. 26 languages or dialect zones were collected, including most of the EU languages
  • LC-STAR - A European Commission-funded project of lexicon development for speech recognition, synthesis, and speech-to-speech translation.
NSC, in its participation in such prestigious consortia, has developed the ability and expertise to produce linguistic resources at a world-class level.

We have experience in the collection of spoken database resources, having collected resources in a diverse range of languages. For example: Hebrew, French, Levantine Arabic, American English, Mexican Spanish, and Korean. In addition, in-house developments of large-scale lexica were developed for the LC-STAR project in American English, French and Hebrew.

In this regard, NSC has the capability of producing linguistic resources at the highest standards. NSC is open to support any specific languages that might be required by its customers. NSC can also provide a targeted collection in order to create customized acoustical models for performing high-accuracy recognition in a specified environment.

© 2010 Natural Speech Communications Ltd. All Rights Reserved.