Loquendo Releases ASR 7.8

Loquendo has released version 7.8 of its speech recognition engine – Loquendo ASR – offering the following enhancements:

• Word lattices can now be exported, opening up to the application layer the possibility of accessing the innermost functionalities of the engine, thus enabling a more accurate recognition of natural, spontaneous speech. The creation and handling of word lattices has also been optimized.
• X-SAMPA phonetic transcriptions are now handled natively by the engine. Furthermore, the Phonetic Learner Tool now generates transcriptions in the X-SAMPA phonetic alphabet; the corresponding lexicon files, generated automatically, are also in X-SAMPA.
• The time needed to build an Adapted Acoustic Model has been dramatically reduced thanks to the upgrading of the Loquendo Acoustic Model Adaptation (AMA) Tool to take advantage of Intel's MKL libraries. The AMA Tool allows acoustic models to be adapted to a particular speaker or to a specific environment (e.g. factory noise, in-car noise) by means of an adaptation process carried out using audio collected in the field.
• Recognition accuracy for U.K. English has been improved by enriching the acoustic models, using additional transcriptions from a custom lexicon.

Loquendo ASR is multilingual, noise-robust speech recognition that currently powers services that handle millions of calls every day across the globe, such as fully automated directory assistance, voice portals, and automotive applications. Loquendo ASR provides specialized acoustic models for telephony (fixed, mobile, VoIP) and for noisy in-car environments, with tools for collecting and automatically analyzing field data. Other features include barge-in for high reactivity and robustness to background speech, and Garbage Clearance for discarding arbitrary spoken sequences not modelled by the grammar (“Um”, “Let me think”, etc.).

Like all Loquendo products, Loquendo ASR is available for both Windows and Linux, while Loquendo Embedded ASR supports all major embedded operating systems.

Loquendo ASR is available in the following 25 languages: U.S. English, U.K. English, Australian English, French, Canadian French, American Spanish, Argentinian, Chilean, Mexican and Castilian Spanish, Catalan, Valencian, Galician, Russian, German, Portuguese, Brazilian, Italian, Dutch, Polish, Turkish, Greek, Danish, Finnish and Swedish, as well as Mixed Language Castilian Spanish and Catalan.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues