Configuration file

Field

Description/Notes

Optional

Default Value

Type

Possible Values

version

Version of the whole document

✘

String

2.0

csdk

✘

Object

csdk/paths

✔

Object

csdk/paths/cache

Absolute or relative to vsdk.json

✔

cache

Path

csdk/paths/data_root

Absolute or relative to vsdk.json

✔

.

Path

csdk/paths/acmod

Absolute or relative to data_root

✔

acmod

Path

csdk/paths/asr

Absolute or relative to data_root

✔

asr

Path

csdk/paths/clc

Absolute or relative to data_root

✔

clc

Path

csdk/paths/clc_ruleset

Absolute or relative to data_root

✔

clc

Path

csdk/paths/dictionary

Absolute or relative to data_root

✔

dictionaries

Path

csdk/paths/search

Absolute or relative to data_root

✔

ctx

Path

csdk/paths/sem3

Absolute or relative to data_root

✔

ctx

Path

csdk/paths/users

Absolute or relative to data_root

✔

users

Path

csdk/paths/audio_based_classifier_model

Absolute or relative to data_root

✔

abc

Path

csdk/paths/confusion_dictionary

Absolute or relative to data_root

✔

dictionaries

Path

csdk/paths/language_model

Absolute or relative to data_root

✔

lm

Path

csdk/tts

✘

Object

csdk/tts/channels

✘

Object

csdk/tts/channels/<channel_name_1>

Name of the channel, used in code

✘

String

csdk/tts/channels/<channel_name_1>/voices

✘

Array

csdk/tts/channels/<channel_name_1>/voices/0

✘

String

<speaker>,<lang>,<quality>

csdk/asr

✘

Object

csdk/asr/recognizers

✘

Object

csdk/asr/recognizers/<recognizer_name_1>

Name of the recognizer, used in code

✘

String

csdk/asr/recognizers/<recognizer_name_1>/acmods

Recognizers accept multiple acoustic models

✘

Array

csdk/asr/recognizers/<recognizer_name_1>/acmods/0

✘

String

csdk/asr/models

✘

Object

csdk/asr/models/<model_name_1>

Name of the model, used in code

✘

String

csdk/asr/models/<model_name_1>/type

✘

String

static, dynamic,
free-speech

csdk/asr/models/<model_name_1>/file

Compiled model file name, extension is .fcf

✘

File

csdk/asr/models/<model_name_1>/sem3

Compiled semantic model file name, extension is .s3c

✔

""

String

csdk/asr/models/<model_name_1>/settings

✔

Object

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_STREAM_RESULT_MODE

The mode in which intermediate results are displayed during recognition. 1 means partial result are activated

✔

0

Int

0, 1

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_ACCURACY

Trade-off between CPU- load, memory requirements and the obtained accuracy of the search

✔

10000

Int

[100 ; 50000]

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_MAXNBEST

Maximum number of hypotheses returned in a result

✔

3

Int

[1 ; 1000]

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_TSILENCE

Minimum amount of trailing silence, in milliseconds. Use a higher value for non- WUW models

✔

100

Int

[100 ; 10000]

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_IG_LOWCONF

Maximum amount of confidence level that indicates that a spoken utterance is out of grammar

✔

5000

Int

[0 ; 10000]

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_IG_HIGHCONF

Minimum amount of confidence level that indicates that a spoken utterance is in grammar

✔

5000

Int

[0 ; 10000]

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_INITBEAMWIDTH

Init beam width. This parameter affects low- level behavior of the algorithm

✔

2500

Int

[0 ; 10000]

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_TANYSPEECH

Allows the recognizer to stop the recognition process during the trailing AnySpeech state

✔

LH_FALSE

String

LH_TRUE, LH_FALSE

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_NBESTRESULT_SETHIDDENKEYS

When enabled additional information is included on the ASR result that can be used for the FM use case

✔

LH_FALSE

String

LH_TRUE, LH_FALSE

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_ONDEMANDLOADING

Context on-demand loading

✔

LH_FALSE

Int

LH_TRUE, LH_FALSE

csdk/asr/models/<model_name_1>/settings/LH_SEARCH_PARAM_SPEECH_TIMEOUT

Speech duration timeout in milliseconds

✔

0

Int

0, [100 ; 60000]

csdk/asr/models/<model_name_1>/acmod

∗Only for dynamic models. Must match with the one configured on the recognizer that will apply this model

✘∗

String

csdk/asr/models/<model_name_1>/slots

∗Only for dynamic models.

✘∗

Object

csdk/asr/models/<model_name_1>/slots/<slot_name_1>

Name of the slot, used in the code

✘

Object

csdk/asr/models/<model_name_1>/slots/<slot_name_1>/slot

Grammar name is marked after !grammar and slot name is after

!slot in your BNF grammar

✘

String

<grammar-name>#<slot-name>

csdk/asr/models/<model_name_1>/slots/<slot_name_1>/category

✔

normal

String

normal, name, artist

csdk/asr/models/<model_name_1>/slots/<slot_name_1>/allow_custom_phonetic

Setting to true will allow for custom phonetic to be provided for this slot

✔

false

Bool

csdk/asr/models/<model_name_1>/lexicon

∗Only for dynamic models.

✘∗

String

csdk/asr/models/<model_name_1>/lexicon/clc

Used during runtime compilation. Use a language that match the rest of the grammar and the recognizer this model will be applied on

✘

File

csdk/asr/models/<model_name_1>/lexicon/settings

✔

Object

csdk/asr/models/<model_name_1>/extra_models

∗Only for free-speech models. All models for a given language must be listed or the program won't function properly

✘∗

Object

csdk/asr/models/<model_name_1>/extra_models/<name>

✘

File

tnl

✘

Object

tnl/paths

✔

Object

tnl/paths/cache

Absolute or relative to vsdk.json

✔

cache

Path

tnl/paths/models

Absolute or relative to vsdk.json

✔

.

Path

tnl/asr

✘

Object

tnl/asr/recognizers

✘

String

tnl/asr/recognizers/<recognizer_name_1>

Name of the recognizer, used in the code

✘

String

tnl/asr/recognizers/<recognizer_name_1>/audio_buffer_size_seconds

In seconds, so multiply by 16000 to get the actual memory space allocated

✔

10

Int

> 0

tnl/asr/models

✘

Object

tnl/asr/models/<model_name_1>

Name of the model, used in the code

✘

Object

tnl/asr/models/<model_name_1>/type

✘

String

static, dynamic

tnl/asr/models/<model_name_1>/file

Compiled model file name, extension is .snsr

✘

File

tnl/asr/models/<model_name_1>/slots

∗Only for dynamic models

✘∗

Object

tnl/asr/models/<model_name_1>/slots/<slot_name_1>

Name of the slot, used in the code

✘

Object

tnl/asr/concurrent_model

∗Only needed if you want to callRecognizer::setModels().

File name of a concurrent template model, extension is .snsr

✔∗

File

vasr

✘

Object

vasr/paths

✔

Object

vasr/paths/data_root

Absolute or relative to vsdk.json

✔

.

Path

vasr/paths/acmod

Absolute or relative to data_root

✔

acmod

Path

vasr/paths/graph

Absolute or relative to data_root

✔

graph

Path

vasr/log

✔

Object

vasr/log/<logger_name>

✔

Object

*, perf

vasr/log/<logger_name>/level

Level of the debugging information printed

✔

String

info, debug

vasr/asr

Object that contains all the configuration of VASR

✘

Object

vasr/asr/recognizers

List here all recognizer that will be created during run-time

✘

Object

vasr/asr/recognizers/<recognizer_name>

Name of the recognizer, used in the code

✘

Object

vasr/asr/recognizers/<recognizer_name>/acmods

Array containing all acmods used

✘

Array

vasr/asr/recognizers/<recognizer_name>/acmods/0

Acmod file name

✘

String

vasr/asr/recognizers/<recognizer_name>/settings

Object containing optional settings

✔

Object

vasr/asr/recognizers/<recognizer_name>/settings/min_speech_duration

Amount of time (in milliseconds) that speech needs to be detected before triggering a recognition. Note that after a successful trigger the X ms of audio of this value will also be passed on to the recognizer. It needs to be a multiple of 100

✔

500

Int

[100 - 10'000]

vasr/asr/recognizers/<recognizer_name>/settings/min_silence_duration

Minimum amount of trailing silence, in milliseconds. Use a higher value for non- WUW models. It needs to be a multiple of 100

✔

700

Int

[100 - 10'000]

vasr/asr/recognizers/<recognizer_name>/settings/padding_size

Amount of perfect silence (in milliseconds) that will be artificially added to the audio at its beginning and its end. This improves performance at almost no cost. Rarely needs to be modified.

✔

Int

[0 - 1'000]

vasr/asr/recognizers/<recognizer_name>/settings/speech_probability_threshold

Set the sensibility of the VAD. A Lower value means more prone to false positives and higher value to false negative

✔

0.5

Float

]0.0 - 1.0[

vasr/asr/recognizers/<recognizer_name>/settings/left_padding_size

Amount of perfect silence (in milliseconds) that will be artificially added to the audio at its beginning. This improves performance at almost no cost. Rarely needs to be modified. Note that this setting is only for free-speech mode. This setting takes priority over the padding_size one.

Since VASR version 2.0.0 (VSDK-VASR version 4.0.0)

✔

Int

[0 - 1'000]

vasr/asr/recognizers/<recognizer_name>/settings/right_padding_size

Amount of perfect silence (in milliseconds) that will be artificially added to the audio at its end. This improves performance at almost no cost. Rarely needs to be modified. Note that this setting is only for free-speech mode. This setting takes priority over the padding_size one.

Since VASR version 2.0.0 (VSDK-VASR version 4.0.0)

✔

Int

[0 - 1'000]

vasr/asr/recognizers/<recognizer_name>/settings/audio_cache_size

Specify the maximum amount of audio (in milliseconds) that a recognizer will keep in its internal buffer. The more audio it has, the higher memory consumption will be. The audio buffering is used for gapless applications. It needs to be a multiple of 100

Since VASR version 2.1.0 (VSDK-VASR version 4.2.0)

✔

10'000

Int

[0 - inf]

vasr/asr/recognizers/<recognizer_name>/settings/invalid_start_time_strategy

If, when calling the Recognizer::setModels function, you pass as a 2nd parameter an invalid start time (e.g. greater than the amount of audio present in the internal buffer), this setting will define how the Engine will react. With warn_and_clamp, a warning will be outputted and the actual startTime that will be used will be the best one possible given the circumstances. With error_and_unset an error message will be outputted and the model will not be set in this recognizer.

Since VASR version 2.2.0 (VSDK-VASR version 5.2.0)

✔

warn_and_clamp

String

[warn_and_clamp, error_and_unset]

vasr/asr/models

✘

Object

vasr/asr/models/<model_name_1>

Name of the model, used in the code

✘

String

vasr/asr/models/<model_name_1>/type

✘

String

static, dynamic,
free-speech

vasr/asr/models/<model_name_1>/file

Absolute or relative to paths/models

✘

Path

vasr/asr/models/<model_name_1>/recognizer

∗Only for dynamic models. Absolute or relative to recognizer

✘∗

File

vasr/asr/models/<model_name_1>/slots

∗Only for dynamic models.

✘∗

Array

vasr/asr/models/<model_name_1>/slots/<slot_name_1>

Name of the slot, used in the code

✘

String

vnlu

✘

Object

vnlu/paths

Object that contains the different path required by the engine to load specific resources

✔

Object

vnlu/paths/data_root

Base location where the engine will load resources from. If relative, it will be based on the configuration file base path

✔

.

String

vnlu/paths/parser

Directory where nlu models (.vum) will be loaded from. If relative, it will be based on the data_root path

✔

parsers

String

vnlu/paths/log

Directory where log will be placed. If relative, it will be based on the data_root path

✔

log

String

vnlu/log

Object that contains all the logging configuration of VASR

✔

Object

vnlu/log/<name>

Configuration for the logger <name>. Possible logger names are [vasr (the default logger), perf (the performance logger), recognizer:<name> (the recognizer named name), model:<name> (the model named name)]. It is also possible to use wildcard as specific places to configure multiple logger at once. Here are the allowed values: [*, recognizer:*, model:*]

✔

vasr

Object

[vasr, perf, recognizer:<name>, model:<name>]

vnlu/log/<name>/level

Set the minimal log level for this logger. Note that setting this will print the level specified and everything above the level e.g. setting the level to warning will still print error and critical messages. Of course warn and warning / err and error are equivalent

✔

info

String

[trace, debug, info, warning, warn, error, err, critical, off]

vnlu/log/<name>/sink

Set the destination of log messages for this logger

✔

stdout-color

String

[stdout, stdout-color, stderr, stderr-color, file]

vnlu/log/<name>/sink_options

Contains all options specific to the type of sink

✔

Object

vnlu/log/<name>/sink_options/path

(For file sink type only) Set the output file path. If relative, it will be based on the paths/log path

✔

String

vnlu/log/<name>/sink_options/max_size

(For file sink type only) Specify the maximum file size (in kbytes) before rotating the file

✔

Int

[1024 - inf]

vnlu/log/<name>/sink_options/max_file

(For file sink type only) Specify the maximum number of log file that will be kept

✔

Int

[1 - inf]

vnlu/log/<name>/sink_options/truncate

(For file sink type only) If set to true file will be truncate / rotate (depending if the maxsize / maxfile options have been specified or not) when the engine is created

✔

false

Boolean

vnlu/nlu

Object that contains all the configuration of VNLU

✘

Object

vnlu/nlu/parsers

List here all parsers that will be created during run-time

✘

Object

vnlu/nlu/parsers/<name>

Specify the definition of a parser. The name is completely free and will be used in the code to create the Parser object

✘

Object

vnlu/nlu/parsers/<name>/model

Path to the nlu model file (.vum) that will be loaded by the Parser. If relative, it will be based on the paths/parser path

✘

String

baratinoo

✘

Object

baratinoo/paths

✔

Object

baratinoo/paths/data_root

Absolute or relative to vsdk.json

✔

Path

baratinoo/tts

✘

Object

baratinoo/tts/channels

✘

Object

baratinoo/tts/channels/<channel_name_1>

✘

Object

baratinoo/tts/channels/<channel_name_1>/voices

✘

Array

baratinoo/tts/channels/<channel_name_1>/voices/0

✘

String

<speaker>

vtapi

✘

Object

vtapi/paths

✔

Object

vtapi/paths/data_root

Absolute or relative to vsdk.json

✔

Path

vtapi/tts

✘

Object

vtapi/tts/channels

✘

Object

vtapi/tts/channels/<channel_name_1>

✘

Object

vtapi/tts/channels/<channel_name_1>/voices

✘

Array

vtapi/tts/channels/<channel_name_1>/voices/0

✘

String

<speaker>,<quality>

tssv

✘

Object

tssv/biometrics

✘

Object

tssv/biometrics/generated_models_path

Absolute or relative to the program's working directory

✘

Path

tssv/biometrics/background_model_TD

Absolute or relative to the program's working directory

✘

File

tssv/biometrics/background_model_TI

Absolute or relative to the program's working directory

✘

File

idvoice

✘

Object

idvoice/biometrics

✘

Object

idvoice/biometrics/generated_models_path

Absolute or relative to the program's working directory

✘

Path

idvoice/biometrics/background_model_TD

Absolute or relative to the program's working directory

✘

File

idvoice/biometrics/background_model_TI

Absolute or relative to the program's working directory

✘

File

s2c

✘

Object

s2c/speech_enhancement

✘

Object

s2c/speech_enhancement/speech_enhancers

✘

Object

s2c/speech_enhancement/speech_enhancers/<name>

Enhancer name. The name is completely free and will be used in the code to create the Enhancer object

✘

Object

s2c/speech_enhancement/speech_enhancers/<name>/configuration

A JSON object that holds the Speech Enhancement configuration generated by vdk-studio

✘

Object