NLU Speech Recognition
Natural dialogs with NLU
CreaVoice® Speech
Recognition
Highlights
- With state-of-the-art natural language understanding (NLU) technology
- Phoneme-based, continuous recognition and word spotting
- Syntax analysis for entire sentence contexts
- Optimized dialog concepts and grammars by many years of experience
CreaVoice® Portals achieve a constantly high recognition performance.
With the help of sophisticated recognition software and a CreaLog-developed syntax it is not only possible to recognize individual words but also entire sentence contexts. Thus complex input options can be described clearly and efficiently, enabling an effective conversational dialog system.
For speech recognition, the digitized speech data of the caller is broken up into its frequency parts in the voice computer. This broken-up data is then compared with the words of the activated dictionary. In the dictionary, the words are filed as rows of phonemes. So the speech data of the caller is continuously compared to these phonemes. The voice portal recognizes a word if the order of phonemes complies with the order of a word in the dictionary.
Apart from the phoneme-based recognition, the speech recognition also uses full-word models which describe entire words, as for example numbers, via their frequency curve. Thus the recognition probability for such frequently used words can be increased tremendously.
With the syntax developed by CreaLog the accuracy of the speech recognition is improved even further. This syntax extends the pure phoneme or word-based recognition with logical alternatives. Thus even entire sentence structures can be described.
With Barge-In - also named Talk-In, Talkover or Cut-Through - everything the caller says, even during the recorded messages of the CreaVoice® Portals are being played, can be recognized and reacted to accordingly. Dialogs thus become faster and the navigation easier and freer.
Technology partners:
|
