GreenKey Technologies, creator of an artificial intelligence (AI) and voice-driven collaboration tool for financial market participants, has announced that the U.S. Patent and Trademark Office has accepted its application for human speech-to-text and subsequent interpretation of spoken financial trades in a fast and accurate manner.
The speech engine, “GreenKey Scribe,” is trained on terabytes of financial conversations and allows for human error correction to achieve industry-leading transcription accuracy. Scribe uses proprietary decoding techniques to transcribe audio up to 10 times faster than other engines and includes natural language processing libraries called Product Class Interpreters to automatically extract relevant data from a conversation.
The tool was designed to help firms comply with MiFID II, which requires the capture of all communications and orders intended to lead to an execution, even if they do not result in a trade. This is especially difficult if the quoting and trading takes place over the phone. With GreenKey’s patent-pending speech-to-text solution, quotes and trades can be extracted directly from historical or real-time conversations and then fed into any execution, analytics or compliance system.
Anthony Tassone, Chief Product Officer of GreenKey Technologies, said: “Scribe learns the nuances of each spoken instrument from the error corrections provided by the user, ultimately enabling accuracy levels as high as 98 percent. By voice-enabling the trader’s workflow, you can maintain a complete audit trail of all voice conversations searchable by quote as well as voice-populating trade tickets or internal systems up to 10 times faster than typing.”
Tejas Shastry, Ph.D., Vice President of Data Science for GreenKey Technologies, said: “GreenKey Scribe combines deep-learning speech recognition, speaker boundary detection and natural language processing to deliver accurate transcriptions with rich, well-formatted metadata. Our multi-faceted approach and focus on both speed and accuracy has made Scribe adaptable to many different workflows.” Scribe is fully containerized for either hosted or on-premise deployments, and audio can come from any third-party phone, trading turret or historical file as well as our own GreenKey Voice platform. Scribe is truly end-point agnostic and designed for the regulated enterprise.”