Using NLP to see

Tejas Shastry

June 4, 2019

There’s no doubt that bank and asset manager trading desks are looking for a way to structure and leverage the overwhelming amount of data generated each day from phone calls, messaging and chat apps, and emails.

GK has historically built and optimized natural language processing (NLP) technology to automatically identify and extract quotes and trades from voice conversations. There are two pieces to this process: speech-to-text (automatic speech recognition or ASR) capability, and text to data (NLP) capability.

Increasingly, significant conversations in trading and banking environments also take place over chat software, which may include Bloomberg, Symphony or other proprietary messaging systems. In order to provide a complete picture of trading opportunities and data, GK’s NLP can now be used to identify and extract quotes and trades from chat sources as well as voice. This is a powerful development because very few solutions can use the subject matter expertise of an NLP “product interpreter” across multiple conversational channels.

GK’s product interpreters, in particular, have been developed specifically for complex financial jargon. They’ve been trained to identify a broad range of references to products across fixed income and equities, including U.S. Treasurys and munis; agency, sovereign, emerging market and corporate bond debt; and U.S. equities including single stocks, futures, swaps and options.

This same finely-honed technology, first designed for phone calls, can now “see” the same quote data in a written message that it “hears” in voice conversation.

For example, here are two ways the same quote information could appear in both chat and voice (spoken) sources:

In both cases, GreenKey’s NLP can detect and tag the same ticker, maturity date, direction, and price referenced in each piece of data, in real-time. Each of these tags, among others, is customizable and dynamic.

Critically, because both chat and voice streams use the same GK product interpreter source library, complementary references to trading data in each can be identified. In other words, when GK’s NLP technology recognizes a particular example from either the chat or voice stream, it recognizes the same example in the other channel.

Not only does this streamline and organize disparate chat and voice workflows, it also allows for new use cases, such as a chatbot that can be queried for quotes, or a product mention in a chat that can trigger a search through your previous voice conversations.

All of this results in a robust accounting of each and every customer interaction, question and possible trade opportunity that may have previously gone unnoticed or uncategorized. Finally, as shown in this video, GK’s Conversation Dashboard allows firms to quickly make sense of this data to understand how to better optimize their business.

Contact GK’s sales team to learn more about how our chat technology can revolutionize your workflows.