SupTech workshop

35 experienced staff in presence (with knowledge and skills enabling to discuss and share experiences).

Prior experience with programming languages such as Matlab, Python, R, or Java is required. Familiarity with natural language processing (NLP) tools and models would be desirable but not considered necessary.

4 April 2023

09:00-09:30 | Welcoming remarks

09:30-11:00 | Overview of SupTech work: EBA, EIOPA, ESMA

11:00-11:15 | Coffee break

11:15-13:15 | From text to data: NLP and other text-processing tools

11:15-12:35 | ESMA project 1: Natural language processing of prospectuses
The focus of this session is on how text-processing tools can allow the extraction of key information from prospectuses and the assessment of unexplored policy-relevant themes.

12:35-13:15 | ESMA project 2: Quantifying potential greenwashing in investment fund
ESMA experts will present text-mining methods used to build a set of indicators to support supervisors in ranking and prioritising possible greenwashing cases for further manual inspection.

13:15-14:15  | Lunch/networking

14:15-16:15 | From text to data: case studies

14:15-15:15 | ECB /SSM tool (a platform using NLP technology to speed up textual analysis)
The ECB will present a recently implemented supervisors’ tool for textual analysis allowing to assess a large volume of textual data, in various formats and within a short time. With an intelligent search engine at its core, the tool employs NLP techniques to help supervisors find, extract, and compare information for their assessments, while combining supervisory data sources in one web-based platform.

15:15-16:15 | Empowering the Smart Supervisor with advanced analytics (DNB)
The DNB experts will present its concept of the ‘Smart Supervisor’ for 2025, i.e. supervisors trained as data literate professionals and fully equipped to take effective supervisory actions based on insights, predictions and recommendations using advanced analytics. This session will introduce two examples of SupTech data products.

16:15-16:30 | Coffee break

16:30-17:30 | ESMA compliance monitoring and data analytics techniques in supervision of data reporting infrastructures (TRs/DRSPs)
At this session ESMA experts will provide an overview of data analytical techniques on big data used in the context of supervision of transaction data reporting infrastructures. The presentation will include real-life examples of tools and techniques used in day-to-day supervision of TRs/DRSPs and provide simple framework to choose the right tools/techniques (including AI/ML) for various analytical problems.

19:00-21:00 | Evening networking programme

5 April 2023

09:00-11:00 | ESMA credit rating agencies supervision : SupTech tools
This session will provide ESMA insight on identification of abnormal data patterns in credit rating action sequences and their classification based on abnormality score and rating characteristics similarity. It includes transitioning from a rule-based to a machine learning-based system.

11:00-11:15 | coffee break

11:15-13:15  | Case studies: ML and AI techniques in supervision; approaches to collecting data and using parsers

11:15-12:15 | An example of the use of ML @ AI techniques in supervision (Banque de France)
At this session the presenter will demonstrate how approximate the stochastic Asset-Liability Management model of a mock life insurer in order to compute its Best Estimate of Liabilities in a Solvency 2 context.

12:15-13:15     Simple Ad-hoc Data request Framework and Code-as-data (BaFin)
The presentation will focus 1) on a hassle-free yet universal data collection scheme; and 2) on “code-as-data” reflecting on the simplicity and wide applicability of parsing. While NLP is essential in retrieving structured information from unstructured sources, parsing can de-mystify programming code and transform it into something more human-digestible.