Explainable Hybrid Artificial Intelligence for Transfer Pricing [TaxoLogic]
Project Description
FFG Projektnummer: FO999936923
eCall Antragsnummer: 66571492
Ausschreibung: AI Ökosysteme 2025: AI for Tech & AI for Green
The project “TaxoLogic” tries to bring computer-support to decisions in tax law, concretely in the field of transfer pricing. We follow a hybrid-AI approach, where
- large language models are used to translate regulatory text (laws, guidelines, etc.) formulated in natural language into a formal language, namely (first order) predicate logic or some variant of it and
- symbolic (automated) reasoning is applied to verify decisions and explain the logical reasoning behind these decisions.
Project Lead
Project Duration
01/03/2026 - 28/02/2029
Publications
2026
[Windsteiger]
Reasoning over Legal Texts Using Large Language Models and Automated Reasoning
Verena Praher, Endre Szasz-Revai, Wolfgang Windsteiger
In: SCML-2026: International Conference on Symbolic Computation and Machine Learning - Extended Abstracts, Bruno Buchberger, François Charton, Matthew England, Cezary Kaliszyk, Manuel Kauers, Hiroshi Kera, Temur Kutsia, Bernhard Moser, Markus Schedl, Wolfgang Schreiner, Martina Seidl, Wolfgang Windsteig (ed.), RISC Proceedings on Symbolic Computation and Machine Learning 3, pp. 76-77. 2026. SCML, https://scml.risc.jku.at/, ISSN xxxx. [doi] [pdf]@inproceedings{RISC7246,
author = {Verena Praher and Endre Szasz-Revai and Wolfgang Windsteiger},
title = {{Reasoning over Legal Texts Using Large Language Models and Automated Reasoning}},
booktitle = {{SCML-2026: International Conference on Symbolic Computation and Machine Learning - Extended Abstracts}},
language = {english},
abstract = {This presentation describes an ongoing project in the area of tax law. The goal of theproject is to bring computer-support to the decision process regarding taxation for internationalbusinesses. In particular, in a first prototype, the focus lies on transfer pricing,which is a non-trivial price-determination and taxation process that concerns businesses with divisions distributedover different countries or economies. We describe the ideaspursued in the project, we do not yet have results nor can we report on their criticalevaluation by practitioners.},
series = {RISC Proceedings on Symbolic Computation and Machine Learning},
number = {3},
pages = {76--77},
publisher = {SCML},
address = {https://scml.risc.jku.at/},
isbn_issn = {ISSN xxxx},
year = {2026},
editor = {Bruno Buchberger and François Charton and Matthew England and Cezary Kaliszyk and Manuel Kauers and Hiroshi Kera and Temur Kutsia and Bernhard Moser and Markus Schedl and Wolfgang Schreiner and Martina Seidl and Wolfgang Windsteig},
refereed = {no},
length = {2},
url = {https://doi.org/10.35011/risc-proceedings-scml.3}
}
author = {Verena Praher and Endre Szasz-Revai and Wolfgang Windsteiger},
title = {{Reasoning over Legal Texts Using Large Language Models and Automated Reasoning}},
booktitle = {{SCML-2026: International Conference on Symbolic Computation and Machine Learning - Extended Abstracts}},
language = {english},
abstract = {This presentation describes an ongoing project in the area of tax law. The goal of theproject is to bring computer-support to the decision process regarding taxation for internationalbusinesses. In particular, in a first prototype, the focus lies on transfer pricing,which is a non-trivial price-determination and taxation process that concerns businesses with divisions distributedover different countries or economies. We describe the ideaspursued in the project, we do not yet have results nor can we report on their criticalevaluation by practitioners.},
series = {RISC Proceedings on Symbolic Computation and Machine Learning},
number = {3},
pages = {76--77},
publisher = {SCML},
address = {https://scml.risc.jku.at/},
isbn_issn = {ISSN xxxx},
year = {2026},
editor = {Bruno Buchberger and François Charton and Matthew England and Cezary Kaliszyk and Manuel Kauers and Hiroshi Kera and Temur Kutsia and Bernhard Moser and Markus Schedl and Wolfgang Schreiner and Martina Seidl and Wolfgang Windsteig},
refereed = {no},
length = {2},
url = {https://doi.org/10.35011/risc-proceedings-scml.3}
}
