Open topic with navigation
IDOL Speech Server can use language models for several languages at a time. The number of languages that you can configure depends on your license. You must configure each language in the
[Resources] section of the tasks configuration file.
[Resources] configuration section lists details for the configured language packs, primarily for use with speech-to-text. This section also contains subsections where you can configure resource packs for language identification, audio fingerprinting, phonetic phrase match, and audio security event detection.
To configure language packs
speechserver-tasks.cfg) in a text editor.
[Resources]section or create one if it does not exist.
List each language that you want to configure in the form
Nis the zero-based rank order for the language.
Typeis the resource type, which is either
sidbase(see Supported Resources). You can omit the type for
langresources, because it is the default type.
LanguagePackis the code for the language pack. For a list of available language packs, see Available language packs.
[Resources] 0=ENUK 1=ENUS 2=langvt:ENUS-pm
Pack parameter to the name of the language pack. For example:
PackDir parameter to the directory where the language model resides. For example:
If the parameter omits the directory path or specifies a relative path, IDOL Speech Server automatically prefixes the value with the default language pack directory path when it searches for the pack.
SampleFrequency parameter to the sample frequency of the audio that the language pack is processing. For example:
If you configure the language pack to use a custom language model, set the
CustomLM parameter to the name and weight of the custom language model separated by a colon (:).
You must also set
CustomDct parameter to the name of the custom dictionary, without the
.dct.sz file name extension. For example:
In this example, IDOL Speech Server uses the file
news.tlm in the custom language model directory with a weight of
0.6 and the dictionary file
news.dct.sz in the same directory.
If you want to change the DNN file to use (for example, to use the smaller, faster DNN provided in the language pack instead of the standard one), set the value of the
DNNFile parameter to the DNN file that you want to use.
HPE recommends (and for 7.0+ versions of language packs, it is compulsory) that you include the following lines in the configuration file for the
[normalizer] modules, so that IDOL Speech Server can access the header to determine the quantity and nature of the extracted acoustic feature vectors:
DNNFile = $stt.lang.DNNFile DNNFileStd = $stt.lang.DNNFileStd
For more information, see the IDOL Speech Server Reference.