2.1.2-alpha.1

Enhancements

• Client Data provided with each request
  • used for troubleshooting, x-client-request-id
• Ability to provide simplified values in truth file testing prebuilt entities
  • only works for CALENDARX at this time
• New validation process up front for truth file
• Refreshed report design
• Improved error visibility
• Latest versions of ASR and NLU protos
• Compiled wordset support for ASR and NLU (reference only)

ASRaaS functionality and params exposed

• utterance_detection_mode
• asr_settings_uri
• discard_speaker_adaptation
• formatting_scheme
• formatting_options
• wakeup_word + filter_wakeup_word
• use_speaker_profile

NLU parsing and eval functionality

• use_entity_literal
• predicted_entities_to_ignore
• entity_evaluation
• nBest
• Confidence thresholds

See config.json for information on each of these.

Fixes and improvements

• Fix issue where report output was not being properly sorted by error rate
• Fix issue with dlm_weight not being properly picked up from config.json
• Fix issue where adjudication map was not properly loaded
• Fix utterance detection mode (default: DISABLED)
• Fix missing logs per request in table
• Minor updates to Test Run Settings tab in .html report:
  • Include the system path to results file
  • Display the language if the default of eng-USA is used when absent from config.json
  • Pretty-print ASR & NLU settings

Thanks,
Nuance Mix Team