New AI eDiscovery tool “reduces legal review hours by 99 percent” – we speak to Veritone

California AI specialist Veritone has launched a new eDiscovery tool, Veritone Illuminate, which searches, analyses, culls and explores audio and video evidence as well as text evidence. It’s unusual to see facial recognition as part of an eDiscovery package: we speak to Chris Ricciuti, vice president of product management for legal & compliance about the new offering and get some private practice reaction.

Veritone claims that in one instance, Illuminate reduced legal review hours by 99 percent, taking 33,000 hours of evidentiary down to 140 hours and predicted costs of $16.5m down to just $5,000.

Illuminate accelerates discovery and reduces costs by making unstructured data searchable by keywords, faces, objects and more. “The key advantages of the Illuminate platform for law firms is that it enables them to cost-effectively and efficiently make sense of their unstructured data,” Chris Ricciuti, vice president of product management for legal & compliance at Veritone told Legal IT Insider.

“If a customer has, say, 3TB of video footage that needs to be reviewed, the previous process might have included humans manually watching those videos to identify particular videos of interest – an extremely time consuming, error-prone and costly process. In Illuminate, customers can instead quickly identify relevant videos by not only searching for keywords, faces and objects, but also by visually exploring the entities and topics that our text analytics capabilities have automatically identified within their dataset.”

Ricciuti added that Illuminate is built on top of Veritone’s platform for AI, called aiWARE. It therefore offers customers a one-stop shop for many AI-enabled capabilities, including machine transcription, machine translation, face recognition, object detection, voice recognition, speaker separation and text analytics.

“Additionally, these AI-enabled capabilities are sourced both from third parties, such as Amazon and Google, as well as being built in-house by Veritone’s data science team. This means that Veritone Illuminate is future-proofed – as new AI capabilities are created or enhanced, they can easily be added onto our platform. This compares to competitors that have a single capability that is “hard coded” and cannot be easily replaced or added to in the future – and customers get locked in as a result.”

Illuminate also differentiates itself through a capability called Conductor, which blends the results of multiple AI-enabled capabilities to produce the most accurate results. “For example, let’s say a financial services customer has a corpus of phone calls that they’ve recorded from conversations that took place on their equities trading desk and they want to make those recordings searchable. Given their nature, the calls in question might contain industry-specific terminology like ticker symbols and mentions of specific securitised instruments,” Ricciuti explained.

“In this case, the customer might want to run various machine transcription capabilities in Veritone Illuminate – ones that are generally trained to understand common words and ones that are specifically trained to understand industry-specific terminology. With Conductor, they can run all of these transcription capabilities and have Conductor automatically blend the highest confidence portions from each transcript to produce a single, highly accurate transcription for each call recording in the corpus. Without Conductor, the customer would be left with either extremely inaccurate transcriptions or having to run each transcription manually and then blend the results themselves.”

Veritone now plans to add new AI-enabled capabilities including the ability for customers to visually explore relationships between identified entities, Ricciuti said. “For example, a user would be able to visually see that John Smith called Mary Smith three times in the past week, click a button and be presented with the call recordings where that actually occurred.”

While we’re skeptical about the wider applicability of the 99% review time reduction statistic, and would hazard that throwing out those kinds of numbers is unhelpful in managing client expectations, it is interesting to see video search packaged up as part of an eDiscovery offering.

Speaking to Legal IT Insider, Nick Pryor, innovations solutions director EMEA and Asia at Bryan Cave Leighton Paisner said: “Audio is more prevalent within eDiscovery reviews than video but it’s interesting to see an organisation taking facial recognition as part of an eDiscovery package, which I’ve not seen before.”