Configuring MRCP for state tables

Configuration of the Blueworx Voice Response MRCP client state table API is defined in the mst.xml file within directory: $CUR_DIR/ca/MST_dir

This file contains XML definitions of specific engines (TTS resources) and the general configuration of the system. The file is processed when the MST custom server is started. This is also when the domain names specified are resolved to IP addresses. Any change made after MST has started will not be effective until after it has been restarted. If a domain name cannot be resolved on start up, the engine will not be added. A sample file, $CUR_DIR/ca/MST_dir/mst_sample.xml, is supplied for reference.

The mst.xml file contains mapping between engine names (for example, gbtts) and a specific MRCP server (for example, machine.hursley.ibm.com, port 554). The engine names specified in file mst.xml define a TTS resource on an MRCP server.

<?xml version="1.0"?>
<!-- This is the MRCP for State Tables config file -->
<config>
   <!-- engine definitions -->
   <engine name="gbtts" type="tts">
      <server>
         machine.hursley.ibm.com
      </server>
      <port>
         554
      </port>
      <mediaurl>
         media/synthesizer
      </mediaurl>
      <initparms>
         speech-language:en-gb
         voice-age:25
      </initparms>
   </engine>
   
</config>

Valid XML elements:

debug

Defines debug options and supports the following elements:

level

Tracing level of the system. This can either be a number or a predefined trace level string. Contact IBM Support for further information.

trace

Components to trace into the AIX trace channel 1 (WVR trace). Comma separated list of component names. Contact IBM Support for further information.

engine

name attribute defines the engine name, type attribute defines tts

server

Address of the text-to-speech server (mandatory)

port

The port to use on the server (mandatory)

mediaurl

A relative URL specifying synthesis server location. This is appended to address for use with RTSP messages. (optional)

initparms

Header fields to be passed on the initial SET-PARMS MRCP message subsequent to the MRCP session being established (mandatory). This element must contain the speech-language: MRCP header field for use as the default language for the engine.

Table 1. Client to MRCP synthesizer header field parameters for use with Blueworx Voice Response. All header fields are optional.

Synthesizer-header field	Description
jump-target	Position to jump forward or backwards in the text being played, relative to the current position.
speaker-profile	The URI of a configuration file that includes a set of voice parameters such as gender, accent, and so on.
voice-parameter	A set of one or more voice parameters such as gender, accent, and so on.
prosody-parameter	A set of one or more prosody parameters such as volume, speed, and so on.
vendor-specific	Allows vendor-specific attributes and values to be sent to the synthesizer to set vendor-specific parameters. Refer to the vendor's documentation for information.
speech-marker	Marker tag to be inserted in the speech data so that an event can be generated at that point.
speech-language	The code for the language of the synthesizer, for example, en-US.
fetch-hint	URI access properties to be used when resources such as documents are retrieved.
audio-fetch-hint	URI access properties to be used when resources such as speech audio files are retrieved.
fetch-timeout	Specifies the timeout duration to allow for retrieving resources.
speak-length	The position in the text being played up to which speech is to be played, relative to the current position.

Refer to the MRCP V1 specification available at http://rfc.net/rfc4463.html for full details. These parameters can be overridden for a specific session by using the MST_TTS_Config state table.

The kill-on-barge-in header field parameter is not supported by the Blueworx Voice Response MRCP state table API. Do not use it in the <initparms> configuration. For details of how to set the behavior for barge-in using DTMF, see the information for the stop_keys parameter of MST_TTS_Speak.