Audio files may seem like an obvious evidence source for forensic investigations, but how we analyse the data it provides is not as clear cut.
In 2018, Deloitte LLP in the UK and US conducted research on communication surveillance practices at global investment banks.
By collaborating with 15 Tier 1 investment banks and a global investment management firm, we were able to establish the prevailing practices, capabilities, and challenges relating to communications surveillance within the securities industry.
Recognising an Audio Surveillance Opportunity
Market trends depict that most organisations review only 0 to 5 percent of electronic communications, despite annually processing over 10 million messages. This is the result of relying on vendor-based solutions, and not implementing machine-learning capabilities, although they have plans to do so. In particular, most organisations have not fully developed an audio monitoring programme, relying on manual processes.
Through collaboration with these financial institutions, Deloitte established that the development of audio surveillance programs dramatically lags behind similar electronic communications surveillance solutions.
The 16 participating financial institutions utilise 13 varying electronic communications surveillance technology platforms. Only 50 percent of the financial institutions were fully satisfied with their current audio surveillance solution, and it was evident that there was a need for providers who could sufficiently fulfil the demand for natural language processing, transcription, and sentiment analysis capabilities as well as achieve the use of machine learning to improve results.
Similarly, there is a current trend of organisations beginning to replace legacy electronic communications and looking into newer technologies.
Deloitte recognised that a comprehensive audio surveillance solution was in demand across industries. However, if we wanted to help meet that need, it was important to identify the key points that led to the low satisfaction rate of the current solutions.
Fifty-six percent of the participating firms planned on introducing machine learning capabilities to improve results, yet none of the firms currently had a solution which fully utilised machine learning.
Our team recognised this as an opportunity to provide our clients with a new solution. Specifically, we wanted to tune the detection of lexicon rules and thus reduce the volume of alerts. Whilst lexicon rule-based detection is currently the most popular method employed to conduct surveillance, use of random sampling and risk-based reviews can augment the lexicon-based detection and has also been pointed out in FINRA’s regulatory guidelines.
From there, the Deloitte team envisaged a voice and video discovery solution, Audio Discovery powered by TrueVoice. Through the fusion of leading transcript technologies, speaker sentiment analysis, and direct integration within Deloitte’s existing e-discovery platform, Deloitte UK created an enhanced voice and video discovery solution to empower our clients and make them more efficient.
The technology used to develop the tool includes the following characteristics, to help our clients address the issues depicted in our research:
- Integrated Technology. Direct integration of a video and audio player within Deloitte’s Relativity platform.
- Speaker Separation. Speaker separation within the audio transcript to deliver an improved review experience.
- Sentiment Analysis. Greater insights delivered through the detection of positive, neutral, or negative sentiment.
- Flagged Phrases. Search term reports linked to hits within the media files with convenient jump-to-hit functionality.
- Cloud-Based Scalable Technology. Flexible and rapidly scalable system resources to meet client demands.
- AI-Enabled Clustering & Classification. Built-in smart grouping of semantically similar content to streamline the review process.
The additional use of third-party technologies further complements Deloitte’s voice and video e-discovery capabilities through the delivery of a standalone phonetic-search platform.
Phonetic search delivers additional value:
- Where the recording quality is low
- Where the transcription output is poor
- Where the search parameters focus on content which is likely to be excluded from common dictionaries.
The application is designed to deliver some important value-drivers for customers.
- Access to the best audio review technology. Removes the need for significant investment in infrastructure, and associated implementation and configuration costs.
- Speed. Takes advantage of our proven extraction and integration methodologies to index thousands of hours in a day and run targeted keyword searches to highlight relevant conversations.
- Efficiency. Provides a scalable solution to meet varying demands and filters substantial volumes of audio economically by removing silence and performing early case assessment through enhanced analytics.
- Cost-effectiveness. Reduces review costs by 50 to 80 percent by limiting the need for manual listening and transcription.
- Wide range of language support. Forty different languages and dialects are recognised and supported, delivering a one-stop translation and transcription solution.
- Phonetically enhanced accuracy. Advanced phonetic search and transcription technologies achieve optimal precise results even across various dialects and accents.
The Nuts and Bolts
The flexibility of Relativity’s open architecture allows various customisations to be built within its data management platform, encouraging innovation and app creation. Thus, we were able to develop Audio Discovery powered by TrueVoice on top of Relativity.
Relativity’s ability to support custom pages and the availability of Application Programming Interfaces (APIs) in its build environment allow for developers and users to communicate with the system to perform bespoke workflows and embed code. This allows teams to create commands and alternative user interfaces, unique to a data type. As a result, Relativity’s open architecture has enabled the creation, development, and integration of Audio Discovery on its widely used platform.
After building the app and submitting it for consideration in the 2020 Innovation Awards, we were also honoured when Audio Discovery went on to win the Solution Provider category for that year. The recognition emphasised the excitement for this type of technology in the legal and compliance communities.
A New Use Case: COVID-19 Compliance
In the midst of the COVID-19 government restrictions, call centre monitoring was incorporated into Deloitte’s Communications, Monitoring, and Surveillance offering in order to help individuals and organisations during these unprecedented times.
Many businesses have enacted business continuity planning, introducing remote working contingency plans. Call centre monitoring can often provide valuable insights to help mitigate operational and financial risks associated with remote working such as compliance with statutory regulations, brand exposure, customer vulnerability, and fair treatment of customers—covered by PRIN (FCA), amongst other regulations. Audio Discovery powered by TrueVoice is ready to help clients and organisations who want to get ahead of these requirements.