Loquendo Releases ASR 7.9

Loquendo this week launched Loquendo ASR 7.9 with enhanced recognition performance and a series of new features.
 
What's more, Loquendo ASR is now also available for the Arabic language, while the acoustic models for U.S. English have been improved, resulting in more accurate recognition of names and addresses for enhanced performance in contexts such as voice-activated dialing, customer self-service, and navigation.
 
Loquendo ASR performance has been further enhanced by a new algorithm for computing confidence scores for the whole utterance, so improving recognition-error rejection. The Loquendo N-best algorithm has also been revised for a reduced response time and faster recognition. 
 
The ARPA format is now supported, so that, when creating N-grams, users have a choice for defining probability information, either through an ARPA or through a standard corpus of word transcriptions. Additionally, the X-SAMPA phonetic alphabet is now fully supported for all languages.

Moreover, Loquendo ASR 7.9 allows speech grammars to reference statistical language models, resulting in a more flexible modelling of the dialogue domain, which leads to a more natural user interaction.
 
ASR 7.9 has reduced memory requirements, enabling more concurrent ASR channels to be supported simultaneously, as well as increased flexibility during the use of grammars at runtime, by inserting meta tags into the grammar source, which allows speech recognition behaviour and configurations to be customized for each individual grammar.
 
In addition, the internal Loquendo transcriber can now be disabled, further reducing memory requirements and permitting developers either to use their own transcribers or to use X-SAMPA transcriptions directly.
 
Like all Loquendo products, Loquendo ASR is available for both Windows and Linux.