Automação do processo de due diligence: como algoritmos podem cortar tempo e trazer eficiência às auditorias legais?

• Autor(a) principal: Oliveira, Daniel Lopes da Silva Ferreira
• Data de Publicação: 2018
• Idioma: por
• Título da fonte: Repositório Institucional do FGV (FGV Repositório Digital)
• Texto Completo: http://hdl.handle.net/10438/24743
• topic: Machine learning; Automation; Legal due diligence; Computational algorithms; Algoritmos computacionais; Auditoria jurídica; Automatização; Clientes; Debêntures; Escritórios de advocacia; Fusões e aquisições; Ineficiência; Clients; Direito

The present work has the objective of studying and demonstrating that the current modus operandi of legal due diligences carried out by law firms has inefficiencies, which can be mitigated by the use of computational algorithms. For this, a Machine Learning prototype that uses n-grams and is based on the Bayes Theorem was developed. Its purpose is to show how it is possible to automate the initial work of mapping and identifying relevant contractual clauses in a legal audit of debentures’ deeds of issuance written in Portuguese, which is not done today. For the Total Issuance Value clause, the algorithm reached, on average, the F1 Score of 77.18%. In the Clause of Early Maturity by Change of Control, scored 74.41%. Finally, when exposed to the Clause of Early Maturity by Cross Default, the F1 Score of the prototype was 72.352%.