Intelligent Voice is pleased to announce the release of Intelligent Voice version 5.4.

This is a major new release containing improvements and bug fixes

Improvements

Punctuation improvements in English mono transcription

A major improvement to English transcript punctuation

INT-5122 Punctuator Model Retrain for Mono file multiple Speakers in an utterance

New container: intelligentvoice.azurecr.io/jumptoweb:1.0

JumpToWeb is Intelligent Voice's standalone web application providing server administration features and acting as a reference and showcase for IV features.

This image is built from drupal:7-apache-bullseye and runs Apache httpd server, serving the JumpToWeb 1.0 web application which uses the Drupal framework

JumpToWeb version 1.0 replaces the version shipped with IV 5.3, and has the following updates:

• New clean install mechanism to bootstrap new installs, including specifying the admin password before install
• Update Drupal (to 7.9.1), drush (to drush/drush:8.4.6), net_gearman (to 2.0.1), PHP (to 7.4.30), and Apache httpd (to 2.4.54)

Note that the IV installer will continue to install Apache httpd which will default to listening on port 1443 instead of 443.  This provides web services for Tag Clouds and People Maps.

New container: intelligentvoice.azurecr.io/sphinxsearch:2.0

Sphinxsearch search engine.  Provides search functionality for the IV search APIs, and lookup data for the Tagger

This image is built from ubuntu:20.04 and runs Sphinx 2.2.11-id64-release (95ae9a6)

There are no functional updates from the version deployed with IV 5.3

New container: intelligentvoice.azurecr.io/gearman-server:1.0

Gearman job server, used by Intelligent Voice as the default queue provider (Azure Service Bus is supported as an alternative).

This image is built from ubuntu:20.04 and runs gearman 1.1.18+ds-3.1ubuntu1

There are no functional updates from the version deployed with IV 5.3

New options for importing from JumpToWeb

INT-5720 Options for selecting Triton models on the Import page

New API for processing timestamp records

INT-5773  Upgrade the processing of timestamp records, to record JSON as either a single item or an array of recordings.  The timestamps are now ISO 8601 with time zone included.

VAD upgrade

Upgraded to 4.0

INT-5886 Dependency on new Get Processing Item API

INT-5736 Upgraded Python 3.7 to 3.8

ASR upgrade

Upgraded to 7.0

INT-5886 Dependency on new Get Processing Item API

INT-5695 Update ASR Punctuator to latest Triton

INT-6256 Missing sseg fix

INT-6140 Keras and Tensorflow unused libraries removal

INT-6081 Fix for disappearance of Beam Alpha/Beta properties

INT-6101 Beam Alpha/Beta properties from global to per-model

INT-6124 Support for all new language identification models

INT-5674 Python 3.6 to 3.8 upgrade

Diarization upgrade

Upgraded to 5.0

INT-5886 Dependency on new Get Processing Item API

INT-5736 Upgraded Python 3.7 to 3.8

Tagger upgrade

Upgraded to 4.0

INT-5886 Dependency on new Get Processing Item API

LexiQal upgrade

Upgraded to Sentiment 4.0, Credibility 3.0, Credibility Score 3.0

INT-5886 Dependency on new Get Processing Item API

INT-5694 Update Credibility to latest Triton

INT-5693 Update Sentiment to latest Triton

DeleteItemWorker upgrade

Upgraded to 3.0

INT-5886 Compatibility with new timestamp API

Triton Installer upgrade

Upgraded to Triton Installer 1.4

New Triton version 22.04

Breaking Changes

All workers must be upgraded

The API for workers has changed (see below).  New versions of every worker are available.  Running the main installation process will upgrade all the core workers but optional workers such as Sentiment, Credibility, DeleteItemWorker etc must also be upgraded following their install guides.

API change

INT-5886 all values previously available from the /properties API endpoint starting starting processed_by_  or sent_to_ are now available from a new /processing endpoint.

New entries are all JSON as either a single item or an array of recordings.  The timestamps are now ISO 8601 with time zone included.

Service name changes

IV services have changed as follows:

sphinxsearch → docker-sphinxsearch

gearman-job-server → docker-gearman-job-server

The installation guide has instructions for how to alias the services under the previous names.  This will prevent breaking of scripts or runbooks.  Future releases of the installer will include these aliases by default.

New Service

JumpToWeb

Restarting the apache2 (Ubuntu) / httpd (RHEL) service now only affects the API methods using Apache httpd.

schema change

INT-5886 all entries in the evitem_properties table starting processed_by_  or sent_to_ have been moved to the evitem_processing table.  Migration script in the installer will move existing data during upgrade.

This enables the processing table to enforce JSON data type (instead of freeform string).  In the evitem_properties, the timestamps are in server local time without a recorded time zone offset.  In the new evitem_processing table, the timestamps included a time zone component.

Bug fixes

INT-2661 Redacted ASR text is still available as a Topic and alternatives in exported .html SmartTranscript

INT-4628 waveforms are only produced for MP3, WAV, FLAC, or Ogg Vorbis

INT-5015 Import Report page times out with lots of data

INT-5731 Silent AVI video file breaks import

INT-5825 orphaned person records in the database after items are deleted

INT-5920 CUDA Linux Repository Key Rotation issue

INT-5946 Tag Cloud and People Map API methods accessible without authentication

INT-5949 Error exporting results for very long files with NASRv3

INT-5953 Groups can be deleted even if items are still in process

INT-5965 Video SmartTranscript requires FROM and TO fields

INT-5966 Lexiqal SmartTranscript error if VideoOCR did not find any text.

INT-5983 Import fails if modelId and modelIds are not specified

INT-5986 Model full name not returned correctly in item API endpoint

INT-5987 Model full name not returned correctly in import list API endpoint

INT-5988 Model full name not returned correctly in import details API endpoint

INT-5989 Docker not starting after reboot (RHEL7 only)

INT-5990 MariaDB repo update (RHEL7 only)

INT-5994 ffmpeg log output files added to import dir

INT-6008 ASRWorker performance impact due to network filtering (Workaround available)

INT-6013 Change default timeout from 30s to 120s for JumpToWeb API calls

Known Issues

INT-2059 Groups API and JumpToWeb groups page timeout with more than 1 million items.

INT-2473 Installer should set file modes correctly regardless of umask

INT-2396 Create new group in the front end does not work if there are no existing groups

INT-2228 JumpToWeb Automatic Topics and People Map order changes on searching

INT-2430 Connections to gearman server are not encrypted (workaround available)

INT-3677 "Data too long for column 'email' " message on creation of email items

INT-3749 Limited options for password characters

INT-4575 Connections to Triton are not encrypted

INT-4970 CrackerWorker not available in 5.3 or 5.4 release

INT-5088 "ERROR: Unknown database 'languages'" at end of installation

INT-5676 Red Hat installer cannot use existing repositories

INT-5698 Connections to the sphinxsearch server are not encrypted (workaround available)

INT-6020 Warning about installing composer as root during install/upgrade

TaggerWorker still uses Python 3.6

Component versions