Recognizer parameter categories

This topic describes groups of related parameters. For an alphabetical list, see the Recognizer parameter reference.

If you use the Management Station to set defaults on a specific Nuance recognition service instance, your settings override the default values.

Application developers do not typically work with recognition server configurations files; they use session.xml files instead.

Licensing parameters

Set licensing parameters during installation:

Parameter	Description	Default
swiep_license_ports	Number of licenses to check out during initialization of the speech detector (endpointer).	8 (licenses)
swiep_license_ports_overdraft_thresh	Number of licenses the endpointer can claim before the system generates a warning.	-1 (disabled)
swirec_license_ports	The number of licenses to check out during Recognizerinitialization.	4 (licenses)
swirec_license_ports_overdraft_thresh	The number of licenses Recognizer can claim before the system generates a warning.	-1 (disabled)
swirec_licensing_features	Specifies which features in the license file to enable at the start of each session.	(all features enabled)

Fetching and caching parameters

Use these parameters to control fetching and caching. Also, see Understanding grammar caching.

Parameter	Description	Default
swirec_disk_cache_enabled	Enables (or disables) the disk and inet caches.	1 (enabled)
swirec_disk_cache_low_water_mark	Largest size of the disk and inet caches after cache cleanup.	400 (MB)
swirec_disk_cache_min_entry_size	Minimum size of disk and inet caches entries.	0 (KB)
swirec_disk_cache_size	Desired maximum size of the disk and inet caches.	500 (MB)
swirec_full_optimization	Default optimization level for fully-optimized grammars.	9
swirec_hits_before_full_optimize	Number of previous activations before fully optimizing a grammar.	3
swirec_inet_user_agent	Specifies the user agent name presented to the web server when HTTP requests are made.	OpenSpeechRecognizer/1.0
swirec_lock_preload_grammars	Keeps preloaded grammars in the memory cache.	0 (flushing allowed)
swirec_memory_cache_low_water_mark	Maximum size of the memory cache after removing grammars to create available space.	0 (adaptation enabled)
swirec_memory_cache_min_entry_size	Minimum size for memory cache entries.	85 (MB)
swirec_memory_cache_size	Desired size of the memory cache.	0 (KB)

You can use these recognizer parameters to set up a proxy, but most deployments use the more encompassing parameters server.inet.HTTPproxyRules and server.inet.HTTPSproxyRules.

Parameter	Description	Default
swirec_inet_proxy_server	Proxy server to be used by Recognizer when fetching grammar URIs.	<value/> (empty value)
swirec_inet_proxy_server_port	The port of a proxy server.	<value/> (empty value)

Optional. Use these parameters to set up mutual authentication when fetching grammars with HTTPS. Most applications use one-way authentication, which does not require that any Recognizer parameters be set.

Parameter	Description	Default
swirec_inet_ssl_ca_certificates_file	A file containing one or more sequential PEM-encoded public CA certificates, used by Recognizer when mutual authentication is required for fetching grammars.	(none)
swirec_inet_ssl_certificate_file	The PEM-encoded client certificate used by the Recognizer when mutual authentication is required for fetching grammars.	(none)
swirec_inet_ssl_private_key_file	The PEM-encoded private key for Recognizer when mutual authentication is required.	(none)
swirec_inet_ssl_verify	Enables peer authentication when mutual authentication is required.	0
swirec_inet_ssl_verify_depth	Limits the depth of the certificate chain for validation when using mutual authentication.	2 (accommodates one intermediate CA)

Recognizer parameters

These parameter cannot be changed dynamically during runtime operation because they apply to the process as a whole and not to individual recognition events.

Parameter	Description	Default
swiep_audio_media_type	Audio formats that will be supplied to Recognizer.	audio/basic;rate=8000
swiep_license_ports	Number of licenses to check out during initialization of the speech detector (endpointer).	8 (licenses)
swiep_license_ports_overdraft_thresh	Number of licenses the endpointer can claim before the system generates a warning.	-1 (disabled)
swiep_min_bytes_to_process	Minimum amount of audio data processed by the endpointer.	800 (bytes)
swiep_waveform_logging_max_channels	Maximum number of channels to save waveforms (recordings of speech from callers).	-1 (no maximum)
swirec_acoustic_adapt_min_num_utts	Minimum data for updating acoustic models.	(depends on language)
swirec_acoustic_adapt_model_update_time	When to update recognition models with learned statistics.	0000 (midnight)
swirec_acoustic_adapt_num_archive	How long to save old statistics files.	3 (months)
swirec_acoustic_adapt_suppress_acoustic_model_update	Stops self-learning adaptation of recognition models.	0 (updates enabled)
swirec_audio_media_type	Audio formats that will be supplied to Recognizer.	audio/basic;rate=8000
swirec_compute_implicit_root	Activates all public rules in a speech grammar as the root rule.	0 (disabled)
swirec_default_optimization	Default optimization level for grammars.	6
swirec_disk_cache_enabled	Enables (or disables) the disk and inet caches.	1 (enabled)
swirec_disk_cache_low_water_mark	Largest size of the disk and inet caches after cache cleanup.	400 (MB)
swirec_disk_cache_min_entry_size	Minimum size of disk and inet caches entries.	0 (KB)
swirec_disk_cache_size	Desired maximum size of the disk and inet caches.	500 (MB)
swirec_full_optimization	Default optimization level for fully-optimized grammars.	9
swirec_hits_before_full_optimize	Number of previous activations before fully optimizing a grammar.	3
swirec_ignore_grammar_media_type	A media type used during grammar activation. (It replaces the media type fetched from the server.)	0 (ignore fetched type)
swirec_inet_query_delimiters	Defines characters that represent delimiters in URI strings.	; and & (semicolon and ampersand)
swirec_language_translation_table	Specifies a text file that maps language declarations in grammars to Nuance language codes.	(Recognizer language codes)
swirec_license_ports	The number of licenses to check out during Recognizer initialization.	4 (licenses)
swirec_license_ports_overdraft_thresh	The number of licenses Recognizer can claim before the system generates a warning.	-1 (disabled)
swirec_licensing_features	Specifies which features in the license file to enable at the start of each session.	(all features enabled)
swirec_lock_preload_grammars	Keeps preloaded grammars in the memory cache.	0 (flushing allowed)
swirec_max_auto_prons	Number of pronunciations to generate automatically when a word is not found.	1 (pronunciations)
swirec_max_source_grammar_size	Maximum allowed size of grammars that can be dynamically compiled.	-1 (unlimited)
swirec_max_training_grammar_size	Limits the CPU and memory cost of SLMs trained at run-time.	-1 (unlimited)
swirec_memory_cache_low_water_mark	Maximum size of the memory cache after removing grammars to create available space.	0 (adaptation enabled)
swirec_memory_cache_min_entry_size	Minimum size for memory cache entries.	85 (MB)
swirec_preload_file	File that loads grammars during Recognizer initialization.	$SWISRSDK/config/SWIgrmPreload.xml
swirec_save_comp_stats	Writes detailed statistics of Recognizer processing to the call logs.	0 (SWIstats is disabled)
swirec_system_dict_name	Name and location of the system dictionary file.	(system dictionary provided by the Recognizer)
swirec_update_interval	Interval for hot insert loading of acoustic models.	300 (5 minutes)
swirec_update_lockfile	Disables the hot insert feature.	$SWISRSDK/config/update_lock.txt
swirec_waveform_logging_index_digits	The number of digits used in filenames when the Recognizer saves waveform files.	2 (limit of 99 files per session)
swirec_waveform_logging_uniform_name	Assigns the same number in the filenames of related waveforms.	0 (disabled)
swirec_word_lattice_density	Density of the word lattice.	100.0
swirec_word_posterior_pruning	Density of the word lattice.	7.0
swissm_confidence_threshold	Default confidence threshold any application SSMs.	0.0

Here are the remaining parameters set in a recognizer configuration file. The values override the installation defaults. They can also be set via other mechanisms.

Parameter	Description	Default
bargein	Allows callers to interrupt prompts.	1 (enabled)
completetimeout	How long to wait before concluding that a caller is finished speaking.	0 (timer disabled)
confidencelevel	Minimum confidence score. Nuance Recognizer rejects utterances with scores below this value.	0 (all utterances accepted)
incompletetimeout	Duration of silence to determine that callers have finished speaking.	1500 (milliseconds)
sensitivity	Sensitivity of the speech detector when looking for speech.	0.5
swiep_BOS_backoff	Safety margin to ensure that the begin-of-speech is captured.	200 (milliseconds)
swiep_EOS_backoff	Safety margin to ensure the end-of-speech is captured.	350 (milliseconds)
swiep_in_prompt_sensitivity_percent	Controls how loudly callers must speak to interrupt prompts (barge-in) and detect speech.	50 (percent)
swiep_magic_word_max_msec	Maximum duration of a magic word candidate for recognition.	800 (milliseconds)
swirec_language_versions	Version of a Recognizer language to use.	(most recent version of each language)
swiep_magic_word_min_msec	Minimum duration of a magic word candidate for recognition.	200 (milliseconds)
swiep_mode	Sets special recognition modes (such as magic word) in the endpointer.	begin_only
swiep_suppress_barge_in_time	Disables barge-in briefly at the beginning of a prompt.	0 (no delay)
swirec_acoustic_adapt_root	Storage location of self-learning files. Controls sharing of models across the server, tenants, or applications.	(empty)
swirec_acoustic_adapt_suppress_adaptation	Temporarily stops self-learning activities for one or more languages.	(depends on language)
swirec_diag_tags_enable	Controls the logged output by enabling tags in diagnostic log files.	(none)
swirec_extra_nbest_keys	Adds grammar keys to the XML result.	SWI_meaning, SWI_literal, SWI_grammarName
swirec_lmweight	Adds weight to match the dynamic ranges of language and acoustic models.	1.0
swirec_load_adjusted_cpu_ranges	Defines ranges of system activity (idle, normal, and busy) based on CPU capacity.	0, 14, 40, 101 (percentages)
swirec_load_adjusted_speedvsaccuracy	Overrides the automatic detection of CPU load, and forces specific values for parameters that balance speed and accuracy.	on
swirec_magic_word_conf_thresh	Confidence threshold for magic word recognition results.	500
swirec_max_arcs	Maximum number of active FSM arcs.	10000, 5000, 3000
swirec_max_cpu_time	Maximum CPU time used to recognize an answer.	20000 (milliseconds)
swirec_max_dict_prons	Maximum number of pronunciations per word.	8 (pronunciations)
swirec_max_logged_nbest	Number of n-best entries written to the call log.	2 (n-best entries)
swirec_max_parses_per_literal	Maximum number of parses evaluated by Recognizer for a single literal string.	10 (parses)
swirec_max_search_time	Maximum CPU time for the search phase of recognition.	5000 (milliseconds)
swirec_max_sentences_tried	Maximum number of candidates for filling the n-best list.	999999 (sentences)
swirec_nbest_list_length	Maximum number of n-best answers that can be returned.	2 (n-best length)
swirec_phoneme_lookahead_beam	Provides a secondary guide to the Viterbi beam search.	-30, -60, -60
swirec_return_waveform	Returns waveforms in recognition results.	1 (enabled)
swirec.secure_context	Sets security levels for protecting confidential data.	open
swirec_selective_barge_in_conf_thresh	Confidence threshold for selective_barge_in mode.	500
swirec_silence_prune_offset	Limits search paths that end in a silence model during pruning.	56, 56, 56
swirec_secondpass_allophone_mapfile_name	Defines allophone maps for secondpass processing in the Recognizer.	(default mapfiles used)
swirec_secondpass_global_fsm_name	Defines finite state machines for secondpass processing in the Recognizer.	(default fsm files used)
swirec_secondpass_model_name	Acoustic models for secondpass processing in Recognizer.	(default models used)
swirec_state_beam	Primary guide for the Viterbi beam search.	0, -15, -35
swirec_waveform_begin_silence	How much silence is kept at the start of a collected utterance.	0 (milliseconds)
swirec_word_confidence_enabled	Controls whether Recognizer performs word confidence calculations.	0 (disabled)

Parameters set in VoiceXML applications

Application developers can set any parameter defined in the VoiceXML standard. They can also set Nuance-specific recognizer parameters using the <property> tag, which the voice browser handles as an MRCP vendor-specific parameter.

These parameters are defined in the VoiceXML standard:

Parameter	Description	Default
bargein	Allows callers to interrupt prompts.	1 (enabled)
completetimeout	How long to wait before concluding that a caller is finished speaking.	0 (timer disabled)
confidencelevel	Minimum confidence score. Recognizer rejects utterances with scores below this value.	0 (all utterances accepted)
incompletetimeout	Duration of silence to determine that callers have finished speaking.	1500 (milliseconds)
maxspeechtimeout	Maximum duration of an utterance collected from users.	-1 (no timeout)
sensitivity	Sensitivity of the speech detector when looking for speech.	0.5
swirec.secure_context	Sets security levels for protecting confidential data.	open
timeout	Specifies how long to wait for speech after a prompt ends.	7000 (milliseconds)

The voice browser must pass these parameters to the Speech Server. It reads the parameters from a VoiceXML document, and performs any needed translation for the recognizer. For example, a VoiceXML value of "10s" might need to be passed as "10000". For a list of needed translations, see Implementing an MRCP client.

Parameters set in session.xml

Application developers can define these parameters in a session.xml file when building and tuning applications. For details, see Configuring sessions (session.xml).

Parameter	Description	Default
DefaultLanguage	Preloads a language during Recognizer startup, and sets the default for built-in grammars.	default (first language installed)
swirec_acoustic_adapt_root	Storage location of self-learning files. Controls sharing of models across the server, tenants, or applications.	(empty)
swirec_acoustic_adapt_suppress_adaptation	Temporarily stops self-learning activities for one or more languages.	(depends on language)
swirec_builtin_grammar_full_dtmf_mode	Includes "*#ABCD" in DTMF built-in grammars.	0
swirec_diag_tags_enable	Controls the logged output by enabling tags in diagnostic log files.	(none)
swirec_enable_robust_compile	Ignores missing pronunciations during grammar compilation.	0 (disabled)
swirec_ignore_grammar_media_type	Ignores the media type that is returned by the server upon fetching a grammar.	1 (ignore fetched type)
swirec_default_language	Default language for built-in grammars.	(value of Recognizer’s DefaultLanguage parameter)
swirec_language_versions	Version of a Recognizer language to use.	(most recent version of each language)
swirec_lmweight	Adds weight to match the dynamic ranges of language and acoustic models.	1.0
swirec_load_adjusted_speedvsaccuracy	Overrides the automatic detection of CPU load, and forces specific values for parameters that balance speed and accuracy.	on
swirec_max_arcs	Maximum number of active FSM arcs.	10000, 5000, 3000
swirec_phoneme_lookahead_beam	Provides a secondary guide to the Viterbi beam search.	-30, -60, -60
swirec_secondpass_allophone_mapfile_name	Defines allophone maps for secondpass processing in Recognizer.	(default mapfiles used)
swirec_secondpass_global_fsm_name	Defines finite state machines for secondpass processing in Recognizer.	(default fsm files used)
swirec_secondpass_model_name	Acoustic models for secondpass processing in Recognizer.	(default models used)
swirec_retain_grammar_import_separator	Retains the semicolon as separator in the query string when importing a grammar.	0
swirec_sensitive_query_keys	Suppresses logging of confidential values in grammar URI strings.	(empty)
swirec_state_beam	Primary guide for the Viterbi beam search.	0, -15, -35
swirec_word_beam	For Nuance use only.

Parameters set in grammar files

Certain parameters can be set with the <meta> tag inside grammar files. This raises a question of when to apply values: during grammar compilation (which allows different grammars to have different values for the same parameter) or during recognition (which requires a single, shared value when different grammars reference each other.

When a parameter is set by more than one active grammar, there are implications for precedence. See Precedence of parameters set via <meta> tags.

Compilation-time parameters

The settings of “compilation-time” parameters are determined when the grammar is compiled. The setting is used in the grammar even if the grammar is subsequently imported by another grammar that sets the same parameter differently.

Parameter	Description	Default
swirec_compile_parser	Speeds recognition time at the cost compilation performance.	0 (feature is off)
swirec_first_pass_grammar	Specifies an n-gram grammar file that defines a Statistical Language Model (SLM).	(empty)
swirec_fsm_grammar	Specifies a finite state machine (fsm) used by a speech grammar.	(empty)
swirec_fsm_wordlist	Specifies a wordlist used by a speech grammar.	(empty)
swirec_max_dict_prons	Maximum number of pronunciations per word.	8 (pronunciations)
swirec_multiword_replace	Limits the number of pronunciations for phrases in user dictionaries.	0 (pronunciations for whole phrases and their individual words)
swirec_optimization	Optimization level for the grammar.	6 (for dynamic compilations)
swirec_normalize_to_probabilities	Improves accuracy by adding a normalized, probabilistic language model.	0 (normalization is off)
swirec_training_grammar	Specifies an SLM training set.	(empty)

Recognition-time parameters

When “recognition-time” parameters are set via a <meta> tag in a grammar, and the grammar is subsequently imported by another grammar, the <meta> setting is ignored and the parent grammar determines the parameter value.

Parameter	Description	Default
incompletetimeout	Duration of silence to determine that callers have finished speaking.	1500 (milliseconds)
swirec_acoustic_adapt_suppress_adaptation	Temporarily stops self-learning activities for one or more languages.	(depends on language)
swirec_app_state_tokens	Adds application or browser information to call logs to synchronize runtime activities with log analysis.	(empty)
swirec_astar_max_paths	Maximum number of nodes visited during the a-star search.	100000 (nodes visited)
swirec_lmweight	Adds weight to match the dynamic ranges of language and acoustic models.	1.0
swirec_max_arcs	Maximum number of active FSM arcs.	10000, 5000, 3000
swirec_max_cpu_time	Maximum CPU time used to recognize an answer.	20000 (milliseconds)
swirec_max_logged_nbest	Number of n-best entries written to the call log.	2 (n-best entries)
swirec_max_parses_per_literal	Maximum number of parses evaluated by Recognizer for a single literal string.	10 (parses)
swirec_nbest_list_length	Maximum number of n-best answers that can be returned.	2 (n-best length)
swirec_phoneme_lookahead_beam	Provides a secondary guide to the Viterbi beam search.	-30, -60, -60
swirec_return_waveform	Returns waveforms in recognition results.	1 (enabled)
swirec.secure_context	Sets security levels for protecting confidential data.	open
swirec_silence_prune_offset	Limits search paths that end in a silence model during pruning.	56, 56, 56
swirec_simple_result_key	Specifies a single key to return in the recognition result instead of all keys.	(empty)
swirec_state_beam	Primary guide for the Viterbi beam search.	0, -15, -35
swirec_word_confidence_enabled	Controls whether Recognizer performs word confidence calculations.	0 (disabled)

Parameters set in parameter grammar files

A parameter grammar sets recognition parameters on all active speech grammars. For parameter grammar format and activation, see Understanding parameter grammars.

These parameters can be set via parameter grammars:

Parameter	Description	Default
completetimeout	How long to wait before concluding that a caller is finished speaking.	0 (timer disabled)
incompletetimeout	Duration of silence to determine that callers have finished speaking.	1500 (milliseconds)
maxspeechtimeout	Maximum duration of an utterance collected from users.	-1 (no timeout)
sensitivity	Sensitivity of the speech detector when looking for speech.	0.5
swirec_acoustic_adapt_suppress_adaptation	Temporarily stops self-learning activities for one or more languages.	(depends on language)
swirec_app_state_tokens	Adds application or browser information to call logs to synchronize runtime activities with log analysis.	(empty)
swirec_barge_in_mode	Sets special recognition modes in Recognizer.	normal
swirec_extra_nbest_keys	Adds grammar keys to the XML result.	SWI_meaning, SWI_literal, SWI_grammarName
swirec_grammar_script	A grammar script to be invoked on the root rule of each n-best result.	(empty)
swirec_grammar_script_sisr	A grammar script to be invoked on the root rule of each n-best result.	(empty)
swirec_load_adjusted_cpu_ranges	Defines ranges of system activity (idle, normal, and busy) based on CPU capacity.	0, 14, 40, 101 (percentages)
swirec_load_adjusted_speedvsaccuracy	Overrides the automatic detection of CPU load, and forces specific values for parameters that balance speed and accuracy.	on
swirec_magic_word_conf_thresh	Confidence threshold for magic word recognition results.	500
swirec_max_arcs	Maximum number of active FSM arcs.	10000, 5000, 3000
swirec_max_cpu_time	Maximum CPU time used to recognize an answer.	20000 (milliseconds)
swirec_max_logged_nbest	Number of n-best entries written to the call log.	2 (n-best entries)
swirec_max_parses_per_literal	Maximum number of parses evaluated by the Recognizer for a single literal string.	10 (parses)
swirec_max_sentences_tried	Maximum number of candidates for filling the n-best list.	999999 (sentences)
swirec_nbest_list_length	Maximum number of n-best answers that can be returned.	2 (n-best length)
swirec_phoneme_lookahead_beam	Provides a secondary guide to the Viterbi beam search.	-30, -60, -60
swirec_return_waveform	Returns waveforms in recognition results.	1 (enabled)
swirec_selective_barge_in_conf_thresh	Confidence threshold for selective_barge_in mode.	500
swirec_silence_prune_offset	Limits search paths that end in a silence model during pruning.	56, 56, 56
swirec_secondpass_allophone_mapfile_name	Defines allophone maps for secondpass processing in Recognizer.	(default mapfiles used)
swirec_secondpass_global_fsm_name	Defines finite state machines for secondpass processing in Recognizer.	(default fsm files used)
swirec_secondpass_model_name	Acoustic models for secondpass processing in Recognizer.	(default models used)
swirec_state_beam	Primary guide for the Viterbi beam search.	0, -15, -35
swirec_waveform_begin_silence	How much silence is kept at the start of a collected utterance.	0 (milliseconds)
swirec_word_confidence_enabled	Controls whether Recognizer performs word confidence calculations.	0 (disabled)
swissm_confidence_threshold	Default confidence threshold any application SSMs.	0.0

Parameters set as environment variables and in SpeechWorks.cfg

Recognizer needs values for some parameters before initialization. These parameters are static and are seldom changed after the initial installation.

Operational parameters

The system administrator sets the following parameters as environment variables or in SpeechWorks.cfg. If the parameter is set in both locations, the environment variable is used and the value in the configuration file is ignored.

Parameter	Description	Default
DefaultLanguage	Preloads a language during Recognizer startup, and sets the default for built-in grammars.	default (first language installed)
GrammarDumpDirectory	Storage location for grammars fetched by Recognizer.	NULL (disabled)
GrammarDumpDirectorySize	Maximum size of the grammar dump directory.	100000 (100 MB)
SWILicenseServerList	Specifies the locations of License Managers.	27000@localhost

Diagnostic parameters in Speechworks.cfg

These parameters are for TRC diagnostic logging. As above, they are set as environment variables or in SpeechWorks.cfg.

Parameter	Description	Default
DiagConfigPollSecs	Defines how often to reload the tagmap files from disk.	600 (seconds)
DiagErrorMap	File that maps diagnostic log messages into any language.	$SWISRSDK/config/SWIErrors.en.us.txt
DiagFileName	Deprecated.	(not applicable)
DiagMaxFileSizeKB	Maximum size of the diagnostic log file.	1024 (KB)
DiagOutputStreamTypes	Writes diagnostic logs to a file, stdout, or both.	debug, file (stdout and file)
DiagSuppressTimestamp	Suppresses timestamps in TRC diagnostic logs.	0 (timestamps written)
DiagTagMapsBaseline	Recognizer’s tagmap files for TRC diagnostic logging.	$SWISRSDK/config/defaultTagmap.txt;$SWISRSDK/config/bwcompatTagmap.txt
DiagTagMapsUser	Application’s tagmap files for custom TRC diagnostic logging.	(empty)

Paths for Recognizer initialization

The Nuance Speech Suite installer sets the following parameter as an environment variable. On rare occasions, a system administrator might change this value.

Parameter	Default	Description
SWISRSDK	Linux: /usr/local/Nuance/Recognizer Windows: C:\Program Files\Nuance\Recognizer	Environment variable pointing to the Recognizer installation directory.

Parameter

Default

Description

SWISRSDK

Linux: /usr/local/Nuance/Recognizer

Windows: C:\Program Files\Nuance\Recognizer

Environment variable pointing to the Recognizer installation directory.

Recognizer parameter categories

Related topics