Skip to main content
  • Call for tenders
  • Closed

Building supervisory reporting concept dictionary through machine learning - FISMA/2018/046/B

A negotiated call for tender to use the machine learning techniques to analyse the given set of legal documents and identify relevant legal concepts – related to the supervisory reporting requirements in particular.

Details

Status
Closed
Publication date
Deadline date
Department
Directorate-General for Financial Stability, Financial Services and Capital Markets Union

Description

Details

Status
Closed

Reference
FISMA/2018/046/B

Deadline
8 August 2018

Description

The European Commission intends to launch a negotiated call for tender in order to carry out a proof-of-concept exercise to perform an analysis of a limited set of regulatory legal documents related to financial supervisory reporting applying natural language processing and machine-learning techniques.

The objective of the exercise is to use these techniques to analyse the given set of legal documents and identify relevant legal concepts – related to the supervisory reporting requirements in particular. The extracted concepts should be organised in a dictionary with cross-references to the related legal text and where applicable also to the related supervisory reporting templates.

The proof-of-concept exercise should consist of an analysis of the legal documents performed by a computer to discover the relevant keywords and concepts (e.g. counterparty, derivative, hypothetical capital) and collect them in a dictionary. The computer algorithm should also establish links between entries in the dictionary and the relevant parts of the legal documents that either define or reference them. Where applicable it should also establish links between concepts and corresponding fields in the structured reporting templates.

An illustrative example of a potential entry in the dictionary:

  • concept: counterparty side
  • defined in: COMMISSION IMPLEMENTING REGULATION (EU) 2017/105, Article 3a
  • reported in: COMMISSION IMPLEMENTING REGULATION (EU) 2017/105, Annex, Table 1, Field 14

Subsequently, the concepts in the dictionary should be categorised into relevant domains (e.g., banking, insurance, transactions, etc.) and organised in hierarchies from the most general to most specific (e.g. contract, derivative, option, etc.).

The final part of the exercise consists of accuracy and scalability assessment. The level of accuracy of the automated concept identification, cross-referencing and categorisation should be measured. The scalability of the approach to the full body of legal documents relevant for supervisory reporting should be assessed and the efficiency of the automated analysis should be compared to a manual analysis.

Because the exercise requires application of state-of-the-art methods, the European Commission may award the contract to two separate tenderers in order to achieve wider coverage of solution approaches.

Contact

Please Note: This overview is for information only. If you are interested in participating in a DG Financial Stability, Financial Services and Capital Markets Union negotiated procedure for low or middle value contracts, you may email your details to ec-fisma-contractsatec [dot] europa [dot] eu (ec-fisma-contracts[at]ec[dot]europa[dot]eu) with a reference to the call for tenders you are interested in.

Your details will be stored in a database for the purpose of tendering only. Inclusion of your contact details does not automatically lead to invitations to tender.