MCC: How has the universe of e-discovery data types evolved?
Patel: Today, in addition to traditional ESI file formats, such as emails and electronic documents, we are seeing text messages and other forms of mobile communications. In the last few years, audio data has become a significant source of e-discovery, which has led to increasing client demand for affordable and defensible methods to manage and review this data format. FTI has always been in the forefront of offering solutions for handling evolving data types in complex e-discovery cases, so developing an audio discovery solution was a natural fit to our existing line of products and services.
MCC: Why should companies care about audio files?
Patel: It's a fact that two-thirds of today’s globally generated data is from audio and video. Also, with growing awareness of technologies that capture and analyze information, people have become very cautious about leaving a paper trail. They often prefer picking up the phone over sending a letter or an email. As a matter of regulatory compliance, some companies are required to record conversations and preserve audio data for three to five years. Given the amount of confidential and discoverable information buried in voice messages and phone recordings, audio data has become a very critical source of information for e-discovery. Therefore, it is imperative for a company to control the generation, preservation and management of its audio data.
MCC: What audio types can be processed through FTI’s Ringtail audio discovery software?
Patel: Audio files can be generated from numerous sources: voicemail, call centers, corporate telecom systems, or even squawk boxes on trading floors, to name just a few. Each organization will have a proprietary system with its own specific configuration for preserving and recording these data types. For e-discovery purposes, files are converted into conventional formats, such as WAV or MP3. Ringtail can handle them all.
MCC: Does this include video recordings?
Patel: Ringtail can indeed process video, but the core data types are different and require a unique workflow.
MCC: How did legal teams review audio files before this kind of technology was available?
Patel: As you can imagine, with the variety of data formats and sources, there was no consistent or standard method for audio review. Lawyers would simply listen to hours of tapes – one at a time and with frequent interruptions to rewind and listen again if the tape is unclear – with the goal of locating information relevant to the case at hand. As such, the judicial system was heavily dependent on an inefficient approach that, in turn, critically depended on each reviewer's knowledge of the case.
MCC: What preservation and discovery requirements are audio recordings subject to?
Jenkins: The 2006 Amendments to the Federal Rules of Civil Procedure (FRCP) and the Dodd-Frank Act tell us that audio recordings are both discoverable and subject to regulatory oversight. So, generally speaking, audio recordings are subject to the same preservation and discovery rules as other types of ESI. But while legal teams have become familiar with “traditional ESI” – office documents and email – audio and other nontraditional sources, perhaps from social media, represent new territory. Here, organizations don't necessarily know what their responsibilities are. And in a recent case – Compass Bank v. Morris Cerullo World Evangelism – a court issued sanctions against the defendant for failing to preserve key audio recordings.
MCC: What are the challenges in dealing with audio discovery?
Patel: Audio has the same processing challenges as the traditional data formats. What do you have? What metadata are you dealing with? For audio files, you can add in factors like sound quality and whether the right compression was used. Given the many unknowns, there's a heightened level of anxiety in dealing with audio files.
Jenkins: And there are additional challenges simply in locating the data and then processing and producing what are often very large audio files. Consequently, legal teams really haven’t approached audio e-discovery with the same intensity as with traditional office-based ESI. From our perspective at FTI, this means that clients need our counsel in getting to the audio that is relevant and should be reviewed.
Patel: Once you start the review, there are new challenges. People may speak in unfamiliar accents or use industry-specific jargon, so you can't take a one-model-fits-all approach when it comes to indexing audio. Phonetics is always at play, and sound fidelity has a big impact on what you hear. Finally, there are added costs with audio, not just on the review side but also in managing recording systems.
MCC: How does FTI’s Ringtail software improve the process?
Jenkins: First, on the process side, we wrap audio e-discovery into the familiar discovery workflow that today’s legal teams already use with traditional ESI. Because clients understand the phases of this workflow, we take away the stress of facing a unique and traumatic project, which then enables them to move through the process more efficiently. But the magic really takes off when you look at how the software handles audio files.
Patel: Ringtail can index and analyze audio data as it does with other electronic files. Obviously, audio involves a phonetic-based searching, so the algorithm for indexing is different, but the process is similar. As a result, Ringtail makes audio files searchable, reviewable, redactable and producible, just like traditional ESI. Now, reviewers can filter results, search for relevance and really perform all functions they would use with emails and e-docs.
Reviewers are comfortable with traditional electronic data because there is an element of consistency in the written language. With audio, however, the same word can be spoken in different ways by different individuals. There could be disturbances and background noises. Ringtail resolves this by indexing a text version of the known words and phonetic information for the words or sounds that don’t have a recognizable translation. Ultimately it generates transcripts from audio files, essentially a text version of your audio, that are then indexed and made available along with traditional data. This is extremely powerful because now clients can easily identify where similar information resides across all data types, and they can leverage Ringtail's additional features, such as mapping or analytics, to perform efficient reviews, again across all data types. In short, Ringtail facilitates a sophisticated, effective and efficient review process that seamlessly folds in audio files.
MCC: Why is it important to review audio files in context with other data types?
Patel: If you and I were discussing something about, say, pollution and then extended the conversation via email, having a transcript of your audio makes it possible to combine both communications and get the full picture.
Jenkins: And going a step further, let’s say that I have a similar but completely separate chat with another colleague. Ringtail will bring together those disparate threaded conversations through its concept clustering and indexing capabilities. Then attorneys who may be aware that you and Nimisha spoke, but weren't aware of my conversation, will be able to make important connections. Attorneys always have an idea of what they're looking for, but by looking at data as it is displayed in Ringtail, they are able to discover unexpected and potentially useful facts. That's really special stuff.
MCC: What are Ringtail audio’s most significant features?
Patel: First, as we discussed, it provides a searchable text transcript for each audio file, which allows reviewers to read through the text while the audio is playing. It enables a more focused, consistent review because reading is the more natural function, and it also saves time because the transcripts remove long silences.
Very often with audio, the custodian information is lost. For example, on trading floors, call lines are shared by multiple traders and cannot be tied to a specific individual. Here, we can perform speaker detection based on speech analysis, essentially training the system to recognize the sound or the cadence of a particular person’s voice, and then search all calls with a similar speech profile. This is powerful because now you can easily find relevant custodians from hundreds and thousands of calls without listening to each one of them. The system grades the calls based on a match percentage, and we have found that if you look at calls at or above a certain percent match, you will find exactly that speaker's voice.
Ringtail also can filter call files based on keywords and any preserved metadata. Finally, the software can redact both the native audio and the transcripts for production purposes. This comes in very handy, for instance, when reviewing call-center calls within healthcare or financial institutions, which involve sensitive patient or credit card information.
MCC: Do those capabilities also apply to privilege review of audio files?
Patel: Yes. It opens up all review work flow capabilities, including complex coding, because the audio is accessible to Ringtail as text.
Jenkins: That's really important. This is more than just finding stuff. It's also satisfying production requests. Before a system like this, reviewers had to put on headphones, listen to hours of recordings and parse out what they needed. Ringtail’s capability to index and identify speakers allows quick access to the relevant parts of a conversation, sometimes over the course of months of phone calls. It’s an incredible time saver.
MCC: What about audio files in foreign languages?
Patel: We currently support audio data processing in 21 languages, including Chinese, Japanese and Korean. We also have dedicated indexing models for U.S. and British English accents.
On top of that, Ringtail’s fine-tuning feature can create an index model for a specific industry, meaning it can comprehend the industry’s jargon in the process. Words that a healthcare call-center rep uses are very different from talk on a trading floor. So you get more accurate results because the system is trained to look for jargon tied to your industry. An example is BOI, a commonly used financial term, which most people would recognize in an audio recording as “boy,” but which we can train Ringtail to understand as “B.O.I.” when appropriate.
MCC: New regulations and the burgeoning use of audio files seem to have created a need that FTI has responded to. What’s behind your responsiveness to these changes in the field?
Jenkins: FTI Technology is distinguished by a roster of consultants who, along with powerful features within our Ringtail platform, help clients solve the substantial and unique challenges we’ve been discussing. We’ve been developing our software in-house for 20 years and have stayed on top of every significant development in e-discovery. Our concept clustering, data analytics and visual review capabilities plus our visual predictive coding were all developed as answers to client problems.
The audio discovery software is no different. A particular client came to us with a problem. We solved it and, in that process, discovered that our solution would work for many others. Our clients are trailblazers, and we love being a part of those initiatives.
Patel: To add to that, we use our own technology internally and, therefore, educate ourselves about its capabilities and challenges. And we back it all up with the consulting piece. As users and consultants, we learn from experience and constantly make improvements so we can offer the most refined product to our clients.
Published September 9, 2015.