Downloads
Repository
Documentation
Issues
Project Media
Google Cloud
Microsoft Azure
Amazon
Nuance Mix
OpenAI
SoundHound
IBM Watson
NVIDIA
Verbio
Deepgram
Yandex Cloud
Speechmatics
ASAPP
GoVivace
Voci
Elevenlabs
Platforms
IBM Watson Speech Recognition (SR) Plugin 1.1.0 to the UniMRCP Server (UMS) has been released.
The plugin is based on the following components:
- UniMRCP Server 1.5.0
- IBM Watson Speech to Text API v1
- Libevent 2.1.8
- Rapidjson 1.1.0
The binaries are currently available for the following Linux distributions:
- Red Hat / CentOS 7
- Ubuntu 16.04 LTS
This release fixes several issues related to processing of received results and stopping no-input timer under certain circumstances. The release also adds support for numerous new configuration parameters.
The detailed list of changes introduced in this release follows.
New Features
- Made the name of the built-in speech transcription grammar configurable, which defaults to 'transcribe'.
- Allow 'builtin:speech/$name' and 'builtin:grammar/$name' to be used interchangeably.
- Accept SRGS speech grammars without processing of any rules defined in the grammar.
- If the parameter 'single-utterance' is set to 'true', recognition completes upon reception of the first final result without waiting for the speech [in]complete timeout to elapse.
Fixed Problems
- Fixed processing of received results to properly observe the 'final' attribute.
- When the configuration parameter 'start-of-input' is set to 'service-originated', then no-input timer might have not been stopped upon start of speech.
Configuration Parameters
- Added a new attribute 'single-utterance', which defaults to 'false'.
- Added a new attribute 'skip-unsupported-grammars', which defaults to 'true'.
- Added a new attribute 'transcription-grammar', which defaults to 'transcribe'.
- Changed the default value of the attribute 'speech-incomplete-timeout' from 1000 to 3000.
- Changed the default value of the attribute 'start-of-input' from 'internal' to 'service-originated'.
Miscellaneous
- Changed grammar referencing routine to be, by default, tolerant to malformed and not supported grammars.
Thank you for using UniMRCP.
--
Arsen Chaloyan
Author of UniMRCP
http://www.unimrcp.org
