Voice Biometric Clustering & Call Summarisation: Announcing Intelligent Voice v6.0

Voice Biometric Clustering For Advanced Metadata Attribution 

Current telephony metadata is often incomplete or inaccurate leading to difficulties for speaker attribution and speaker search capabilities. While telephony systems and therefore recording platforms may be capable of capturing a log-in name or telephone number this metadata does not contain any information regarding who was actually speaking on any given audio file, presenting a major attribution, search and retrieval issue, particularly for financial services compliance use cases.  

For compliance the requirement is to find communications in relation to a specific “person”. Searching for a number or log-in does not provide any guarantee of returning all communications in relation a specific individual. This problem is particularly complex with “open mic” systems such as Turrets and Kiosks where a speaker could be any one of dozens of team-mates or counterparties with no metadata attribution.  

Fortunately the human voice and its vocal characteristics show unique qualities,  and these can be used to help to identify a person  thanks to Biometric Identification technology.  The Intelligent Voice labs have been busy developing a further refinement to its Voice Biometric technology, Voice Biometric clustering,  which automatically assign individuals a unique “voiceprint”.  The clustering solution uses state-of-the-art biometrics to passively enrol every speaker across all recorded communications and provides a unique identifier for every speaker. This unique identifier is attributed to every speaker and surfaces, with a high degree of accuracy, the file name and time stamps for everywhere this person speaks within every audio file, even on lower quality calls. This information can then be used to accurately search for a “person” within large audio datasets independently of any other metadata attribution.   

Call Summarisation 

Do you write extensive summaries for a given call or retrospectively listen to a call, in order to catch-up or review for investigation purposes? Then Intelligent Voice’s Call Summarisation feature could save you a huge amount of time with precision accuracy to boot. The Call Summarisation feature automatically identifies the key components of a call and produces a summary of the content within each key interaction, distilling hours of speech down into concise and informative bite sized sentences. The initial release of the feature is available for English only, the models are currently being trained in the lab for further language support.   

The Stack Is Now Fully Containerized  

The team at Intelligent Voice have achieved a major milestone in the release of IV 6.0 in that all of the components of the application have been containerized. With this 100% containerization coverage we can now recommend a fully-automated pipeline to deploy container images, and a staged deployment pipeline to update to new versions. This will enable a number of advantages for our customers: 

– Faster access to latest features – components can now be released in isolation meaning access to new Intelligent Voice features can be accelerated and with minimal effort  

– Easier to scale  

– Improved Efficiency – resource usage lower than with VMs  

– Flexibility – allows customers to integrate with their existing DevOps environment 

– Improved security – regular patch updates and applications can be isolated from the host system and from each other 

In addition to these major improvements, we have introduced a number of smaller enhancements that will add value to you our customers and we are committed to continual iterative development as we continue to bring you cutting edge technology for all your audio, video and chat surveillance needs.  

If you are looking for a demo of current or new functionality, please contact your usual sales representative or sales@intelligent-voice.test

Please get in touch with your support representative to talk about upgrade and deployment options, or contact support@intelligent-voice.test