From unstructured market conversations to a fine-grained understanding of liquidity in OTC capital markets.
From unstructured market conversations to a fine-grained understanding of liquidity in OTC capital markets.
Our natural language models understand the complex jargon used in capital markets enabling extraction of information from across the life cycle of a trade.
A view of what clients are saying rather than what they’ve executed gives you a view of trades that haven’t yet happened, maximizing available liquidity.
We turn trading desks into modern data-driven franchises with state-of-the-art natural language processing that gets to the heart of complex conversations.
Critical to our approach is the use of advanced language modelling techniques that go beyond ‘scraping’ technology. Language modelling has improved dramatically in recent years, but to make them useful in capital markets Sense Street has developed a proprietary pipeline that conditions these models on the domain.
Data scientists, linguists, traders collaborate to create uniquely annotated datasets focused on OTC conversational flow.
Engineers and NLP researchers fine-tune Sense Street’s suite of models, using our proprietary curated datasets.
Our models extract counterparty intentions even when expressed across multiple unstructured conversations.
Sense Street’s processes and systems have been carefully constructed to pass information security concerns typical within the industry, which means pilots can be executed in as little as 3 months.
Our models are trained in a federated manner from industry wide datasets guranteeing high generalization.
In production our solutions can be deployed on premise or as a service depending on integration requirements.
Pilots are executed on ISO27001/SOC 2 certified networks that pass easily stringent impact assessments.