by Deborah Blaxell, Legal Consultant and Martin Bonney, International Director, Consulting Services, Epiq Systems.
Increasing regulation of industry, together with businesses’ growing reliance upon digital devices which enable and record oral communication has combined to create data search and review requirements which cannot easily be satisfied by means of traditional technologies and techniques. Audio evidence, like documentary evidence, can be pivotal to a legal case and a failure to deal with this evidence effectively and efficiently can leave businesses open to judicial criticism, damaging publicity and searing fines from regulators.
During litigation and regulatory investigation, those making or defending legal claims are often required to disclose audio evidence. The definition of “electronic document” in the Civil Procedure Rules includes electronic communications such as voice mail and communication recorded on mobile phones and other electronic devices and media. Litigants are required to identify, preserve and disclose the pertinent parts of their audio evidence, along with other forms of evidence, as soon as litigation proceedings are contemplated, with the threat of adverse inferences being drawn if such evidence is deliberately withheld. Litigators have traditionally shied away from disclosure of audio evidence, on grounds of proportionality, complexity and the perceived inadequacy of existing technical solutions. However, the disclosure of all types of electronic evidence, including telephone recordings, is now very much embedded in the litigation process.
Recent allegations of misconduct and collusion relating to LIBOR rate fixing and phone hacking which came to light in 2011/12, further highlight the value of audio evidence in deterring and detecting market abuse. Many regulatory authorities recognise the compelling nature of audio recordings. In the UK, the Financial Services Authority (FSA) introduced rules in 2008 requiring that all firms regulated by the FSA record all telephone conversations and electronic communications relating to client orders and the conclusion of transactions in the equity, bond, and derivatives markets. In November 2011 this requirement was extended to cover the recording of mobile phone conversations that relate to client orders and transactions by regulated firms. Similar rules have either been introduced or are under consideration by regulators across the globe. Whilst IT departments of regulated businesses have taken technical steps to comply with these obligations, their review systems have been designed around the need to provide a small scale sampling of a particular individual’s calls over a short period of time, rather than a comprehensive and defensible collection over an extended period, as typically required for litigation or a major regulatory investigation.
Unsurprisingly, regulators are increasing their investment in systems which will help them investigate and review audio evidence. For example in the US, the Department of Justice, Criminal Division has
licensed cutting edge audio discovery software to help with their investigations and it is inevitable that others will follow. Litigants and those who work within regulated industries need to maintain awareness of developments and best practices to enable them to monitor and, when appropriate, disclose audio evidence.
Challenges in dealing with audio evidence:
Much discussion has centred upon the challenges which exist for those working with audio evidence and the increased difficulty involved in searching and reviewing this evidence. As with all electronic evidence, the place to start is with information governance and efficient storage of data. Being able to effectively log calls, to provide custodians and telephone numbers on a specific date (and export the data in a consistent and defensible manner), can allow effective triage of data and save time further down the track.
Matters are made complicated by the fact that different business sectors use different modes of telecommunication. For example, trading desks will often make recordings not just of direct telephone calls, but also of “hoot and holler” boxes which involve permanent open circuits between two or more parties conducting multiple conversations, including hours of silence. Thinking about how to collect data and filtering out redundant recordings before review is a significant challenge.
When it comes to review, the traditional method of reviewing audio evidence is by linear review where reviewers – usually lawyers – listen to audio recordings and review it as they listen. The benefits of this method are obvious: the reviewer (often a lawyer) is able to identify speech, understand nuance and meaning and make professional judgements upon relevance.
However, this approach is not scalable. It is not unusual to be faced with the recordings of tens of thousands of hours of conversations. The review of audio evidence is potentially much more labour intensive than other types of electronically stored information because of the need to listen not only to the words but, as mentioned earlier, to consider tone, expression and other subtle nuances of speech and intonation which may hold evidential clues. Regional accents and individual conversation styles add to the difficulty and commentators suggest that it takes approximately four hours of human time to review each hour of audio evidence. The variety of devices onto which voice can be recorded makes retrieving the evidence complex and can affect the quality of the recordings and the speed at which the reviewers can assess the evidence. As a result, the cost of reviewing audio evidence can quickly become prohibitive and proportionate in only the most valuable of cases.
Foreign languages can add further challenges to a review. It is not always possible to source native speakers within the timeframes and budgets available and especially in Europe, there may be multiple languages to review. Identifying foreign language content and allocating the calls to appropriate reviewers is not a simple task, so a high level of quality control is therefore required to produce a repeatable (and hence defensible) process.
A potential solution to the prohibitive cost of the linear review is the conversion of speech to text. Transcribers produce a transcript of the audio, which can then be searched with technology using standard keyword searching techniques. The responsive set can then either be used to identify portions of the audio for lawyer review, or in some cases the transcripts themselves are treated as evidence.
Human transcription suffers from many of the same problems as the linear review. It is slow, costly and languages remain an issue. Decisions about nuance and meaning are taken by non-lawyers unfamiliar with the matter, so crucial evidence may be missed. However, there may be merit in an automated transcription route, and solutions are being developed to make this an effective solution for legal purposes.
The future of audio review
Phonemic search technologies are a relatively new entrant to the world of e-Discovery but are gaining rapid acceptance. Using phonetic audio search engines, forensic data search can be conducted across multiple sound file types to find relevant material extremely quickly compared to traditional methods. Phonetic search technology transforms audio recordings into a phonetic representation, rather than written words. It includes a model for the way in which words are pronounced and is therefore not limited to only searching for words in a dictionary. Audio recordings are then indexed by identifiable units of sound – the building blocks of human speech known as ‘phonemes’. Extensive research goes into building these phoneme algorithms, and they are structured around various languages – so for example US English is slightly different to UK English and Latin American Spanish is different to Spanish spoken in Spain. It is also possible to run several language packs across the same data sets, so that (for instance) one can search for both English and German terms.
Once the recordings are processed in this way, the audio files are available for search and analysis by expert reviewers. This presents a new challenge to lawyers, since it involves a twist on the conventional key word approach. Within text, a given keyword will render a 100% success rate, so long as the word is spelt correctly within the document. However, with sounds, some sounds are simply too generic. As such there is a need to fine tune the process by identifying context. Once the lawyers are confident in a search / filtering strategy, the data can be filtered effectively and potentially relevant documents made available for conventional review via standard review platforms. This process is best undertaken through an iterative approach (new terms may arise while listening) and also a statistically sound sampling of non responsive documents. The benefit of this solution is that it enables a responsive set of audio to be produced from a collection in a streamlined manner which can significantly reduce the burden of traditional linear review.
This is still a new area. The technologies and techniques which are available to deal with audio evidence are evolving rapidly and it is important for lawyers to keep abreast of these developments. Dealing with audio evidence efficiently, such that the cost remains proportionate to the overall cost of the legal exercise, is a challenge. However, it will be increasingly difficult for businesses to complain that audio evidence is too complex to deal with. As ever, while the technologies are crucial to providing a solution, a considered and well thought through plan and consideration of the available options will stand those who are required to search, review and disclose audio evidence in good stead.
 Practice Direction 31B, part (3).
 Many financial institutions across the globe were implicated in a libor rate fixing scandal
 News International staff were found to have intercepted telephone calls to win stories
 The Regulators are already tuned in…why aren’t you? Audio Discovery, 5 July 2012 by Jeff Schleuter.