Speech synthesis and voice recognition are now twice as fast in Voximplant Kit
Starting on December 18, Voximplant Kit will have a new and improved IVR block with faster and more accurate speech synthesis and recognition.
IVR enables you to use voice synthesis or audio recording to speak with your clients. Clients have the ability to interact with the bot both orally (speech recognition) and by pressing keys on the phone (DTMF).
What improvements were made?
ASR recognition now redirects the response to the branch of the IVR block as soon as it determines the correct answer so there is no need to listen until the client finishes speaking. Recognition is now done piece by piece using the interim.result setting. When this approach is ineffective for a particular call; that is, if the keywords defined by the dictionary of the scenario’s block are not found by the system in the client’s speech, then the old asr.result recognition is automatically turned on. asr.result listens to the entire phrase, searches for the trigger word(s), and then routes the call along the desired branch.
The new algorithm minimizes the pause after receiving the desired trigger, as recognition of the response stops immediately after the spoken word coincides with the one specified in the dictionary. On average, long answers containing a keyword at the beginning of the phrase are processed twice as fast than before.
But these are not the only changes. Improvements have affected speech synthesis - TTS synthesis used to voice the text only after it processed the entire phrase but now, the recording is downloaded in parts and after the first part is processed, the robot begins to pronounce the text. Subsequently, the second part is downloaded, and so on until the end of the text. This reduces pauses before starting playback.
As a result of improved ASR recognition and TTS synthesis, IVR Voximplant Kit has become twice as fast and thus, a better option for your business needs.