Applied AI in OTC Derivative Confirmations

Automating complex workflows requires judicious use of both Supervised and Unsupervised Machine Learning techniques

Financial Institutions are struggling to simplify and automate the operations supporting Over-the-Counter (OTC) Derivative trades. Industry reliance on unstructured data - specifically PDFs - adds an additional engineering challenge. 

Artificial Intelligence (AI) and Machine Learning (ML) techniques provide an effective path toward automating the OTC Derivative Trade confirmation process. Combining supervised and unsupervised learning, AI solutions recognize variations in economic and legal terms within “paper” trades, including those based on the International Swaps and Derivatives Association (ISDA) Master Agreement. The unstructured data in confirmation PDFs (designed for human eyes only and often shared person-to-person via email) is data that basic automation like RPA cannot process.

Let’s dive into supervised and unsupervised Machine Learning models and how they apply to the OnCorps Derivative Confirmations application.

Supervised Learning 

Supervised Learning, also known as Supervised Machine Learning, is a subcategory of Machine Learning and Artificial Intelligence. It is defined by its use of datasets labeled by humans to train algorithms, which an then classify data or predict outcomes accurately. (IBM)

??How Supervised Learning is applied within OTC Derivative Confirmations - Powered by OnCorps AI

OnCorps uses supervised learning to extract economic terms automatically from paper confirmations (ISDA and others) sent via email in PDFs. We then match those terms to a client’s internal booking system. 

Document Classification

Document Classification is a supervised Machine Learning model for categorizing documents based on their content. Document classification can help analysts better understand data assets across different mediums, including text-based files and multimedia like audio clips and videos. (MonkeyLearn)

??How Document Classification is applied within OTC Derivative Confirmations - Powered by OnCorps AI

OnCorps uses Document Classification to determine which counterparties are involved, and which products are contained, in a paper OTC confirmation.

Named-Entity Recognition

Named-Entity Recognition (NER) is a Supervised Learning model used for information extraction — enabling the identification and categorization of entities such as people, organizations, locations, and time periods from unstructured text. This subtask of Natural Language Processing (NLP) can be used to gain valuable insight into vast amounts of data while automating tedious manual processes. (Depends on the Definition)

??How Named-Entity Recognition is applied within OTC Derivative Confirmations - Powered by OnCorps AI

OnCorps uses NER to extract economic terms within ISDA confirmations. NER helps identify variances between  PDF confirmations sent from  counterparties, and is one of the first steps toward intelligent analysis.

Unsupervised Learning 

Unsupervised Learning is a type of Machine Learning algorithm that looks for patterns in a dataset without labels provided by humans.  This technique gives systems the ability to make sense out of large data sets without requiring any external guidance. AI algorithms are able to autonomously uncover patterns within a given dataset that enable accurate classification and grouping. (IBM)

??How Unsupervised Learning is applied within OTC Derivative Confirmations - Powered by OnCorps AI

OnCorps uses Unsupervised Learning models for legal checks within PDFs. The system is trained on a set of trade confirmations. New confirmations are reviewed for anomalies, and exceptions are flagged for human intervention. As an additional aid for follow-up, the magnitude of the change is also indicated; small changes may be inconsequential, and larger changes can be prioritized appropriately.

Discuss Supervised and Unsupervised Learning With OnCorps

The OnCorps OTC Derivative Confirmation solution is one example where a host of AI techniques have been assembled to solve an issue too challenging for other technologies.  AI, ML, and NLP are employed for tasks that exceed the ability of simpler automation tools like RPA — automatic ingestion, matching, and routing confirmations into appropriate work queues. 

These technologies underpin a collaborative workflow shared by all counterparties. When confirmations contain changes or outright breaks in legal terms or other details, counterparties can resolve issues all within the OnCorps application, streamlining an otherwise cumbersome process for the middle office. 

Schedule a demo by clicking here.