Asset 1
Digital Reasoning Allowed Significant Patent for using Machine Learning in Electronic Communications Surveillance

NASHVILLE, Tenn., January 25, 2018 – Digital Reasoning, a leader in Artificial Intelligence (AI) for understanding human communications, announced that the U.S. Patent and Trademark Office has allowed the application for a new patent covering the company’s Conduct Surveillance solution. Several related patent applications are pending.

Financial services firms are looking for increased protection against market abuse and conduct risk. Traditional surveillance technologies cannot effectively monitor the exponentially increasing data volumes since they are rudimentary and focus on lexicon-based searches. They are not designed to uncover patterns, people networks, or capture sentiment and determine suspicious behaviors.

Globally, regulators now expect financial institutions to retain and monitor a multitude of electronic communications and perform enhanced supervision, adding non-traditional sources (e.g. social media, voice) to traditional channels (e-mail, trader chat). Financial institutions are being compelled to update their technology and processes and are implementing sophisticated, holistic surveillance programs that can collectively analyze all communications data, trade and order data, and other behavioral data such as badge swipes, printer usage logs, and call data records.

The rapid emergence of technologies like natural language processing (NLP), machine learning, improved voice analytics and evolving leading practices are now forcing firms to reassess their approach to e-communications surveillance.

The allowed patent application, for an invention by Digital Reasoning’s collaborators Dr. Kenneth Graham and colleagues titled “Systems and Methods for Identifying Violation Conditions from Electronic Communications”, covers several capabilities that are critical to effective surveillance solutions:

  • Determining the meaning of words in the context of emails and chats based on the context of how they are used and across multiple communications using machine learning techniques that identify conduct risks
  • Resolving entities and people across multiple communications using context and activities as inputs to machine learning algorithms
  • Flagging behavior patterns of employees based on rules and training by example and providing those alerts in a queue for review
  • Leveraging feedback from that queue to improve the algorithm’s ability to predict potential conduct breaches or policy violations
  • The use of metadata in electronic communications to enable and enhance the detection of the above violations and feedback into the system
  • The development and use of language patterns indicative of violations
  • The development and use of key features such as domain, audience, and tone of communication in these indicators
  • The presentation and workflow presenting these violations to users and capturing their feedback to improve the system’s machine learning models

“Digital Reasoning was the first to apply machine learning to teach large-scale systems how to identify interesting human behaviors in communications data,” said Tim Estes, president and founder at Digital Reasoning. “We are securing protection of our intellectual property that puts AI to work for financial firms that must meet regulatory demands such as MiFID II and MAR and seek Digital Transformation of their business to better serve their clients.”

Digital Reasoning continues to innovate in areas such as Machine Education – the process of teaching computers human domain knowledge through data – as well as in speech analytics using state-of-the-art Deep Learning neural architectures. Some of these innovations are being used by large global investment banks.