Voice Identification Management by Intelligent Voice

Endorsed by NVIDIA at Founder, President and CEO level, Voice Identification Management is available at all stages of the Electronic Discovery Reference Model from Identification & Collection through to Production & Presentation.

Voice Identification Management screenshot

Voice Identification Management correctly attributes the same individual across multiple spoken languages and dialects and is robust for wide variations in audio quality.

Voice Identification Management is deployable prior to, alongside, or post Relativity processing.

Voice Identification Management provides a matrix of names linked to processed recordings in extracted metadata style for upload into Relativity or any downstream Trial Preparation platforms.

Voice Identification Management Functionality

  • Differentiates between individuals on challenging audio such as compressed trader turret recordings, where recording quality might otherwise mean a speaker remaining indeterminate between two or more possible speakers.
  • Review and analyse voice recordings as a single phase exercise, rather than moving back and forth across the case timeline as additional instances where a particular speaker is present on calls are identified by human review.
  • Where an unknown John or Jane Doe is encountered and the contents of the conversation mean any other recordings where they also feature might be of interest, that set of recordings can be rapidly determined and then reviewed.
  • Applicable to what might otherwise by plausible deep fake video or audio recordings.

Benefits of Voice Identification Management

  • Provides robust and defensible identification and collection for voice independent of the quality or otherwise of information governance prior to the case.
  • Mitigates the absence of a regulatory minimum for voice recording quality in many jurisdictions and domains.
  • Review and analyse all recordings where a particular speaker features regardless of metadata attribution.
  • Demonstrably performant even where reviewers, long familiar with accents in a particular locale and with the custodians themselves, struggle to distinguish between different people speaking.
  • Mitigates the full spectrum of voice metadata issues including:
    • Borrowed credentials for desk phone logins
    • Out of date phone number registries
    • Conference call speaker phones
    • Withheld caller details

From Intelligent Voice