After four years of research and development, Nashville-based Digital Reasoning has released an AI-enabled voice analytics solution for financial services communications monitoring. It has also announced a partnership with regulatory risk tech company, Corlytics, to deliver one of the highest volumes of pre-trained models for monitoring available to the financial sector, according to the companies.
Monitoring telephone calls for misconduct has been an ongoing challenge for both financial firms and their regulators. There are many reasons for this. First, financial markets are full of domain specific terms and are populated by individuals speaking English in a wide variety of accents, as well as foreign languages. On trading floors, there can also be quite a lot of background noise. All these factors can create transcriptions with lots of errors.
As well, most solutions try to find suspicious calls based on a set of terms fed to them by the compliance team. So, for voice communications, the rates of false positives is traditionally quite high, because solutions focus on keyword spotting. As a result, compliance teams have struggled to recruit enough people to analyse all of the calls that are flagged.
The advent of deep learning is beginning to change this, says Brandon Carl, vice president of product management at Digital Reasoning. Deep learning enables higher quality voice analytics, including more complex natural language understanding. It can also look at behaviours within language.
While in the past, solutions had to be fed with specific trigger phrases by compliance teams, the Digital Reasoning solution is able to learn itself, assembling uses of language by those potentially engaged in misconduct that compliance teams may not have noticed. The overall result is greatly improved transcripts and lower false positive rates, Carl says.
According to Carl, this approach enables voice and text-based communications such as emails to be analysed together. This potentially reduces false positives, increases the chances of catching misconduct, and allows firms to take a more holistic approach to misconduct monitoring. He says Digital Reasoning has translated all of this into a user-friendly interface and analytics.
He says: “Deep learning reduces the complexity in terms of some of these deployments. The second thing that it enables you to do, is to more quickly adapt to things like national domains, like non-English languages. The third thing, and really one of the most interesting things, is the same technology can be used to do voice classification, event detection, those sorts of things. So, what was previously an entire portfolio of technology, is now compressed into one key technology when used and architected correctly. This sets up a lot of interesting possibilities.”
Partnering with Corlytics
Digital Reasoning is developing in other ways, too. The company is combining its AI communications monitoring capabilities with Corlytics’s analysis of regulatory risk. The companies say this will enable financial firms to more effectively monitor communications channels for market manipulation and other regulatory infractions. Mike O’Keefe, general manager at Corlytics, says: “The ability to use AI to effectively interpret intent and behaviour tied to our regulatory mapping differentiates our offering from legacy solutions that simply focus on lexicon and rule-based approaches.”
Through its partnership with Corlytics, Digital Reasoning’s models have already been mapped to over 42 regulatory provisions covering different components such as market manipulation, insider trading, bribery, complaints handling, fraud, client/investor confidentiality, mis-selling and suitability, disclose-to-market, and arrears handling. The company has a goal of further expanding its regulatory use cases globally over the next year.