Natural Language Processing In Capital Markets

What is NLP? 

Simply put NLP is a way to allow computer to read and interpret data that is very likely unstructured.

This short paper explores how NLP is used in Capital Markets.

The growth of electronic trading across all asset class has led to tools such as NLP becoming increasing popular in decision making in trading.

Electronic trading across asset classes  is nothing new. We are also seeing algo trading being utilised in more asset classes.

• The 2020 Algorithmic Trading Market report,  stated 60-73% of U.S. equity market trades utilise algorithmic trading.
• Refinitiv state FX algo trading as a percentage of FX spot volumes is now at 20 percent. A 2019 survey by JP Morgan, estimated 80% of JP Morgan’s FX flow would come via electronic channels.
• Research by Greenwich Associates found in the corporate bond market on a daily basis one-third of traders execute  21% of their volume using an algo  program. The same report stated electronic trading for in corporate bonds had soared to 34%

Trading volumes during Q1 2020 increased dramatically driven by Covid and other events. For example Stataisa  noted equity volumes globally increased by 34% in Q1 2020 vs Q4 2019.

The key message to take away from this is electronic trading and algo trading are increasing

Algo trading Strategies

For simplicity lets group algo trading strategies into three categories

• Optimal Trade Execution – the algo seeks to  limit the negative price impact of a trade
• Market Making – the algo focuses on making money from the bid ask spread
• Arbitrage – the algo is designed to leverage market imperfections such micro price differences for the same instrument on different trading venues.

The strategies above rely mainly on quantitative data and most of this is easily obtainable, at a cost.

Where would you use NLP?

You would use NLP if you wanted to incorporate un-structured data sets into your algo. These could include

-       Company reports including annual reports and 10K reports

-       Social media feeds including tweets and facebook

-       ESG reports including those published by independent third parties

-       News feeds including Bloomberg, Thomson Reuters, BBC

-       Blogs and white papers including academic papers

-       Research papers including those from research analyst.

By using NLP traders and investors can determine if the sentiment expressed in annual report or ESG reports is positive, negative, or neither etc. The result of this sentiment analysis can then be used to drive trading decisions. While NLP can used to analyse ESG reports. NLP itself can be power hungry and could leave a large carbon footprint.

Another example of NLP providing sentiment analysis is reviewing content on social media platforms such as twitter and facebook.

NLP is also used to analyses transcripts of earning calls to identify signals.  If the earning call is broadcast live the analysis can be taken one step further by using software to analyse the body language of the presenters.

Compliance is also an area where NLP can be very useful. NLP can be used to review emails, chatter via IMs and other communication tools to help identify any potential activity that needs further investigation. NLP can also be used to assist with KYC. Late in 2019 UBS announced a program to explore using NLP for KYC.

The Bank of England is using NLP to measure the complexity of banking reform for the financial crisis. Traders can also use NLP to analyse reports by the Bank of England, regulators and governments. An example is analysing Covid 19 reports produced by governments and other organisation to identify trading signals.

NLP is a very useful tool for the capital markets community. Its use will only increase and the quality of output from NLP will continue to improve.