AWS Transcribe Plugin 1.3.0 Released

[Arsen Chaloyan]

Amazon Web Services (AWS) Transcribe Speech Recognition Plugin 1.3.0 to the UniMRCP Server (UMS) has been released.

The plugin is based on the following components:

• UniMRCP Server 1.7.0
• AWS SDK 1.8.129
• Libevent 2.1.9

The binaries are currently available for the following Linux distributions:

• Red Hat / CentOS 7 (unimrcp-transcribe-1.7.3-1.el7.x86_64.rpm)
• Red Hat / CentOS 8 (unimrcp-transcribe-1.7.3-1.el8.x86_64.rpm)
• Ubuntu 16.04 LTS (unimrcp-transcribe-bot_1.7.3-xenial_amd64.deb)
• Ubuntu 18.04 LTS (unimrcp-transcribe-bot_1.7.3-bionic_amd64.deb)

This release adds support for numerous new parameters and fixes a possible segfault. The release also addresses an issue in the AWS SDK initialization encountered only when the use of the license server is enabled.

The detailed list of changes introduced in this release follows.

New Features

• Allow vocabulary-name to be specified per recognition request.
• Added support for vocabulary-filter-name, vocabulary-filter-method, language-model-name, partial-results-stability, enable-partial-results-stabilization, show-speaker-label.
• Made separator of built-in grammar parameters configurable.

Fixed Problems

• Fixed a possible segfault in processing of a Websocket closure event.
• Reordered the AWS SDK initialization to make sure the SDK is initialized prior to initialization of the licensing client used to connect to the license server, if enabled. Otherwise, the SDK initialization could fail.

Configuration Parameters

• Added new attributes 'vocabulary-filter-name', 'vocabulary-filter-method', 'language-model-name', 'partial-results-stability', 'enable-partial-results-stabilization', 'show-speaker-label' to the element 'streaming-recognition'.
• Added a new attribute 'grammar-param-separator' to the element 'streaming-recognition'.

Miscellaneous

• Updated the Usage Guide to reflect the changes introduced in this release.

Visit the Transcribe plugin page for more information.

http://www.unimrcp.org/transcribe

Thank you for using UniMRCP.

--

Arsen Chaloyan
Author of UniMRCP
http://www.unimrcp.org