Unless you’ve been living in a vacuum, you’ll know that the major banks are severely restricting or banning the use of chat rooms because it’s alleged they were used to collude in the fixing of LIBOR, forex and precious metal prices. The move will likely cause a bunch of headaches, not least for Bloomberg – which maintains the largest chat service, because traders and brokers are so reliant on chat for keeping in touch with eachother and the market, for putting out RFQs and so on.
But as we all know, banning something people like doesn’t necessarily mean that’s the end of it. By hook or by crook, people will find a way…
Using text analytics to monitor conversations for market abuse will be at the heart of any solution that enables financial firms to keep the chat rooms alive.
Bloomberg and Thomson Reuters have already announced services to help Compliance departments monitor chat. I’m sure many banks and brokers with proprietary systems have done, or are doing, similar.
Risks of Multi-Asset Staking and Ways To Solve Them Using CeDeFiGo to article >>
Thomson Reuters and Bloomberg systematically look for mentions of entities – companies, instruments and people – in close proximity to keywords, phrases or events, to generate alerts for Compliance analysts. For example, talking about the “London Fix” and mentioning “EURUSD” on the last day of the trading month or leading up to 4pm London time could trigger an alert to the analysts. That’s basic text analytics – the same kind of technology that’s used to trade based on news. And like trading, they would probably be able to tell the ‘relevance’ of an entity in a chat, ie, is the person mention directly involved, and the ‘novelty’, that is, is this the first time these two parties have talked about this particular event.
But will that be enough to keep the banks, brokers and regulators happy? Probably not – it is likely the owners/managers of chat functionality will have to invest more heavily in semantic or graph search functionality – the same kind of search used by Amazon and Facebook. Amazon uses it to tell you ‘people who bought that gizmo also bought this gadget’, and Facebook to suggest who your friends should be and what adverts to put in front of you.
Semantic search pulls up the relationships between entities. That means the Compliance folk will not only get alerted to potential abuse, but they’ll be able to trace the thread of the conversation through a relationship graph. They won’t just see one or two participants having a conversation, they’ll be able to see the network.
Of course, banks, brokers and other owners of chat rooms will have to prove the technology works. It’ll probably come down to analyzing an historical archive of conversations around a situation we now know was rigged to see if they could’ve prevented any collusion.
It’ll be costly to establish the technology, but likely worth it to keep a very handy tool alive.