PROJECT
completed
More-With-Less — Efficient Language Models for SMEs
MISSION
The core goal of this project
DATEXIS investigates data- and model-efficient methods for adapting and specializing language models to domain-specific tasks with limited training data.
Overview
In the BHT subproject, data- and model-centric approaches are explored for three use cases: automated text authoring for the finance industry, document matching for legal tech companies, and improved reading-order recognition in legal and finance documents.
FUNDING BODY
BMBF · Bundesministerium für Bildung und Forschung, KI in Unternehmen
GRANT ID
01IS23013C
PERIOD
2023-2025
PARTNERS
Merantix Labs GmbH2txt NLG GmbHFraunhofer IAIS
TEAM · CONTRIBUTORS
