Nexidia Brings Neural Phonetic Speech Analytics to Audio Discoveryposted by Remco Witkamp
Bumicom’s partner Nexidia Brings Neural Phonetic Speech Analytics to Audio Discovery Enhancements provide visual data analysis and improved tools for review and workflow for the legal market. Nexidia, a provider of audio and video search technology, has added Neural Phonetics Speech Analytics to Nexidia Search, its discovery application for the legal market, to provide improved workflow and discovery tools for the review of audio evidence. Neural Phonetic Speech Analytics combines Nexidia’s automatic speech recognition engine with its patented phonetic index to produce word-level transcription and sentiment analysis, helping reviewers understand the context of their audio recordings. The system applies Nexidia’s latest advances in neural-net processing and language models for improved performance throughout the application. These features add a new visual approach to enable legal reviewers to automatically see and analyze the most common themes and topics that occur within audio files. The system displays several reports, including word clouds depicting frequency of occurrence at both the word and phrase level, and a related phrase diagram that lets users show the most common words and phrases in relation to each other. This output can then be translated directly into queries that can be run against the full phonetic index to bring up all of the files that correspond to these searches. Neural Phonetic Speech Analytics “is a visual analytics tool” that allows users to see what is showing up in their audio recordings, explains Jeff Schlueter, vice president and general manager of legal markets at Nexidia. “It improves the way we do phonetic indexing, making it faster and giving it a greater degree of accuracy.” Users, he says, “can see what is being spoken in their legal content, and they can then translate that into their search queries.” This latest product release, Schlueter says, adds concept search to Nexidia’s audio discovery solutions, allowing users to start with a visual reference to see what words and phrases show up frequently in the audio and then use this information to search more deeply into the content using Nexidia’s traditional phonetic search. In addition, the word-based analysis can use other textual content, such as email, transcribed depositions, and news articles, to help improve the phrase recognition output and deliver results that closely represent the content of the recordings. “The system can scale to handle any amount of content,” Schlueter says. “There’s really no limit to how much content you can feed into the system.” Nexidia Search also includes improved workflows for assigning files to review teams, enhanced tools for managing large sets of review tags and metadata categories, and enhancements to its Relativity plug-in to support Relativity 9.0. According to Schlueter, Nexidia launched the same technology into its audio search products for contact centers last year. “We’re now taking what we learned in the contact centers and bringing it to the legal market,” he says. The capabilities, he adds, are becoming a core part of Nexidia’s full solution set. All of these new features are targeted for general availability by late March.