AI for Fraud detection in Trading: A Case Study

Those familiar with stock markets surely know how traders communicate with each other via chatrooms in order to keep up with the trends and get actionable insights from peers. Most of these platforms, especially those offering free access, are a cluster of activity that needs to be moderated to ensure the interactions are civil and legal.  Why do it manually through  professional moderators when you can use an automated AI system?

According to Fortune, Wall Street banks were allegedly manipulating the market for US Treasury bonds. There are documented cases where these platforms have been used by some traders to coordinate transactions that are clearly fraudulent. Proof of their guilt can be found in the conversations held in trading chat rooms where traders rigged the LIBOR benchmark interest rate and a sum of benchmark foreign exchange prices over eight years. What if this misbehavior could have been detected faster? The total impact of the fraud would have been significantly lower.

Together with one of the Big Four, Bitext developed a system that monitors trader chat conversations and issues alerts of potentially fraudulent intent. For example, the system flags sentences where the main verb indicates illegal activity (e.g. actions belonging to the semantic field of “set”, “lower, “fix”, etc.) about a financial product or index (as the direct object of the verb, like “Libor rating”, “fixing”, or “Euribor”) in sentences such as “can we lower our fixing today please?”. Bitext’s syntactic analysis identifies these sentences without mixing them with innocent uses of “set” or “fix” and “Libor”, such as “set an appointment to discuss Libor issues”.

The key to this enhanced technology lies in our linguistic approach. Our NLP resources are thoroughly tagged and classified by our linguistics team, leaving little room for error when it comes to semantics and grammar issues. What’s more, the technology can be customized and ready in less than expected. Do you want to see how our solutions could help your business? Schedule a demo with us. 


Subscribe to email Updates