Lay summary

A modern document understanding AI for the legal profession

Eine moderne KI zum Verstehen von Dokumenten für die Rechtsbranche

Ziel des Forschungsprojekts

Document Smartification. DeepJudge kann automatisch wichtige Informationen hervorheben und mit externen Datenquellen wie Gesetzbuchern, Gerichtsurteilen, Handelsregistern und Enzyklopadien verknupfen. Mit unserer KI sind alle Informationen, die Juristen und Juristinnen benotigen, hochstens einen Klick entfernt, womit die Bearbeitung eines Falles deutlich beschleunigt werden kann.

Kontext-sensitive Anonymisierung. DeepJudge kann sensible Informationen automatisch und mit vollem Kontextverständnis anonymisieren. Die Fähigkeit, über den Kontext nachzudenken, hebt unsere KI erheblich von bestehenden Tools ab, die lediglich auf das Vorkommen vordefinierter Kategorien von Wörtern prüfen. 

Gesellschaftlicher Kontext

Recht erschwinglicher machen: Recht und Gesellschaft sind untrennbar miteinander verbunden. Mit der zunehmenden Komplexität der Gesellschaft ist auch die Komplexität der Rechtsprechung gewachsen. DeepJudge steigert sowohl die Effizienz als auch die Qualität der Arbeit von Anwälten/-innen, womit letztlich auch die Rechtshilfe für Bürger/-innen erschwinglicher werden.

Abstract

Legal teams are more overloaded than ever: suffering from slow processes and inefficiencies, they spend a vast amount of time working on tedious low-level tasks that can be automated with AI. We have conducted dozens of interviews with legal experts and are in close contact with several leading law firms in Switzerland and Germany, who have provided us with deep insights into their work routine. The emerging picture is clear: the manual processing of large document collections is one of the biggest culprits in the workflow of lawyers.Paulina is a Co-founder and CEO of DeepJudge AG, founded by four ETH Machine Learning PhDs with over 20+ years of combined experience with Deep Learning and Natural Language Processing at scale. The four of us met during our doctorate at the Machine Learning institute at ETH, where we quickly realised the immense untapped potential of AI-powered document processing tools in the legal world. Our mission is to develop a context-aware legal document processing AI that is a generation ahead of existing legaltech software.

Smartifying Documents. DeepJudge can automatically highlight important information and smartify documents by augmenting them with external data sources such as codes of law, court rulings, registries of commerce and encyclopedias. With our document smartification, all the information lawyers need is at most one click away. A wealth of information at their fingertips!

Context-sensitive Anonymization. DeepJudge can anonymize sensitive information automatically and with full context understanding. The ability to reason about context sets our AI well apart from existing tools that merely check for the occurrence of predefined categories of words.

Our Secret Technology. Trained on gigabytes of legal data, our multi-lingual AI has an unprecedented level of context-sensitive understanding of legal concepts and legal language. Context-awareness is the quantum leap that separates DeepJudge from existing tools. This capability allows DeepJudge for instance to anonymize passages that would reveal sensitive information about an entity (e.g. a person or company) even without that entity being mentioned. Developing an AI with such capabilities is highly non-trivial: the actual learning process we use to train our AI is of-course our well-kept company secret.

The BRIDGE funding will specifically be used to fully develop and monetize the document smartification and document anonymization features, which are currently in the advanced prototype phase (see demo video https://youtu.be/lggv9g4kzG0), into two licensable DeepJudge products that we expect to start selling to law firms and legal teams in Switzerland and Germany by the end of 2021.

Last updated:09.06.2022