IBM Watson Speech Recognition (SR) Plugin 1.5.0 to the UniMRCP Server (UMS) has been released.
The plugin is based on the following components:
- UniMRCP Server 1.6.0
- IBM Watson Speech to Text API v1
- Libevent 2.1.9
- Rapidjson 1.1.0
The binaries are currently available for the following Linux distributions:
- Red Hat / CentOS 7 (unimrcp-watson-sr-1.6.3-1.el7.x86_64.rpm)
- Ubuntu 16.04 LTS (unimrcp-watson-sr_1.6.3-xenial_amd64.deb)
- Ubuntu 18.04 LTS (unimrcp-watson-sr_1.6.3-bionic_amd64.deb)
This release adds support for custom language and acoustic models and also allows to specify base model version.
The release initially adds support for numerous new parameters settable per recognition request such as word-confidence, timestamps, speaker-labels and others.
The detailed list of changes introduced in this release follows.
- Added support for an optional language parameter passed to a built-in grammar.
- Added support for custom language and acoustic models.
- Added support for base model version.
- Added support for certain vendor-specific parameters, including 'speech-start-timeout'. See section 4.7 in the Usage Guide.
- Added support for the content type 'text/grammar-ref-list'.
- Added support for numerous new parameters settable per recognition request to the service such as 'word-confidence', 'timestamps', 'speaker-labels' and others. The parameters can be set globally in umswatsonsr.xml and be specified per recognition request either via vendor-specific parameters or optional attributes passed to a built-in grammar or via metadata set in an SRGS XML grammar. See the Usage Guide.
- Make sure START-OF-INPUT is sent before sending RECOGNITION-COMPLETE with a completion cause set to 'no-match' or 'success'.
- Do not set speech/result flag if the detector is already in the complete state. This could result in an attempt to send another audio chunk, when the input completion was already signaled.
- Compose the header field Waveform-URI based on the protocol version. Before, the format defined in MRCPv2 was used unconditionally.
- Fixed output format of an RDR to strictly conform to JSON.
- Added new configuration parameters 'language-customization-id', 'acoustic-customization-id', 'base-model-version', 'customization-weight', 'word-confidence', 'timestamps', 'speaker-labels', 'redaction', 'processing-metrics', 'processing-metrics-interval', 'audio-metrics'.
- Updated the Usage Guide to reflect the changes introduced in this release.
Visit the Watson SR plugin page for more information.
Thank you for using UniMRCP.
Author of UniMRCP