TeliSpeech

Languages

The teliSpeech Recognizer is used with at least one teliSpeech language. Indeed, telisma provides the teliSpeech language catalogue developed using the Studio Languge Development Kit..

A teliSpeech language is set of language-specific software components and data, for instance; a dictionary, phonetic rules, acoustic libraries.

The main European languages as well as the main Indian languages and a set of Asian, North and Latin American languages are already available to use with the teliSpeech Recognizer and new ones are continuingly being developed. All these languages benefit directly from telisma's background, taking into consideration the wide range of training databases and the variety of lexicons used. Each representation of a language is composed of a set of components which are either inherent to the acoustic representation or to the phonetic definition of this language.

Precise acoustic libraries

Acoustic libraries are the acoustic representation of all the phonemes (sound units) of a language. They are generated from speech data collected from a very large number of speakers. The speakers are selected with a wide range of demographic criteria such as age, sex and geography. The recordings collected from them cover both mobile and landline, quiet and noisy environments. Therefore, even the most complex words can be recognised by teliSpeech!

A complete dictionary

The dictionary for a teliSpeech language contains the pronunciations (and their variants) of many thousands of words from the language, all of these pronunciations have been optimised and benefit from telisma's long experience in the field

Ready to use grammars

The Built-in grammars shipped with each teliSpeech language are standard grammars which have been optimised by telisma. They cover a range of common application components such as numbers and dates.

Thematic lexicons

Thematic lexicons, such as for city names or proper nouns, are included in the basic teliSpeech language. For example, within the teliSpeech French language, even the name of the smallest village has already been included.

Higher quality still

Acoustic libraries enhanced with data collected from your application will raise the performance levels of your service to unprecedented levels. You enter into a 'virtuous circle', and better recognised users are more satisfied users!

Your own lexicons

It is also possible for you to add your own pronunciations of words in user lexicons and therefore recognise application or domain specific vocabulary, such as technical terms, abbreviations, acronyms or brand names.