Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law

Marina Luketina; Lukas Knogler; Christoph Georg Schütz

Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law

Marina Luketina^*
, Lukas Knogler
, Christoph Georg Schütz

^*Corresponding author for this work

Institute of Business Informatics - Data & Knowledge Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference proceedings › peer-review

Abstract

In tax advisory practice, case descriptions are typically not structured in a machine-readable format, with clients describing their situation in natural language. Large language models excel at natural-language understanding. However, for legal reasoning, including tax law, the propensity of LLMs to hallucinate presents a considerable challenge. Rule-based systems, on the other hand, offer verifiably correct reasoning given the correct input. Therefore, in this paper, we propose a hybrid approach to support tax advisors with analyzing tax cases, combining a rule-based system with large language models. We focus on the analysis of chain-transaction cases in value-added tax (VAT) law, where the law states a clear set of rules for regular chain-transaction cases. We employ a large language model (LLM) for the construction of structured representations of natural-language VAT case descriptions and law-based rules for the identification of the movable supply, which determines tax liabilities. Human tax advisors can obtain a graphical visualization of the structured representation to verify the correctness of the LLM's output while the law-based rules return reliable decisions.

Keywords: Neuro-symbolic artificial intelligence, Knowledge graphs, Decision support systems, Tax management, Value-added tax.

Original language	English
Title of host publication	Joint Proceedings of the 16th Workshop on Ontology Design and Patterns and the 1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025), Nara, Japan, November 2-3, 2025
Editors	Fjollë Novakazi , Aryan Singh Dalal
Publisher	CEUR Workshop Proceedings (CEUR-WS.org)
Pages	130-142
Number of pages	13
Edition	1
Publication status	Published - Dec 2025

Publication series

Name	CEUR Workshop Proceedings
Volume	4093

Fields of science

102030 Semantic technologies
502050 Business informatics
102010 Database systems
102035 Data science
503008 E-learning
502058 Digital transformation
509026 Digitalisation research
102033 Data mining
102 Computer Sciences
102027 Web engineering
102028 Knowledge engineering
102016 IT security
102015 Information systems
102025 Distributed systems

JKU Focus areas

Digital Transformation

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

https://ceur-ws.org/Vol-4093/Paper5hai.pdf

1 Participating in a conference, workshop, ...
1 Contributed talk

Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law
Schütz, C. G. (Speaker) & Luketina, M. (Speaker)
02 Nov 2025
Activity: Talk or presentation › Contributed talk › science-to-science
1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025)
Schütz, C. G. (Participant)
02 Nov 2025 → 06 Nov 2025
Activity: Participating in or organising an event › Participating in a conference, workshop, ...

Cite this

Luketina, M., Knogler, L., & Schütz, C. G. (2025). Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law. In F. Novakazi , & A. S. Dalal (Eds.), Joint Proceedings of the 16th Workshop on Ontology Design and Patterns and the 1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025), Nara, Japan, November 2-3, 2025 (1 ed., pp. 130-142). (CEUR Workshop Proceedings; Vol. 4093). CEUR Workshop Proceedings (CEUR-WS.org). https://ceur-ws.org/Vol-4093/Paper5hai.pdf

Luketina, Marina ; Knogler, Lukas ; Schütz, Christoph Georg. / Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law. Joint Proceedings of the 16th Workshop on Ontology Design and Patterns and the 1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025), Nara, Japan, November 2-3, 2025. editor / Fjollë Novakazi ; Aryan Singh Dalal . 1. ed. CEUR Workshop Proceedings (CEUR-WS.org), 2025. pp. 130-142 (CEUR Workshop Proceedings).

@inproceedings{bd556871a3364cf1b8fbc036bb7026c6,

title = "Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law",

abstract = "In tax advisory practice, case descriptions are typically not structured in a machine-readable format, with clients describing their situation in natural language. Large language models excel at natural-language understanding. However, for legal reasoning, including tax law, the propensity of LLMs to hallucinate presents a considerable challenge. Rule-based systems, on the other hand, offer verifiably correct reasoning given the correct input. Therefore, in this paper, we propose a hybrid approach to support tax advisors with analyzing tax cases, combining a rule-based system with large language models. We focus on the analysis of chain-transaction cases in value-added tax (VAT) law, where the law states a clear set of rules for regular chain-transaction cases. We employ a large language model (LLM) for the construction of structured representations of natural-language VAT case descriptions and law-based rules for the identification of the movable supply, which determines tax liabilities. Human tax advisors can obtain a graphical visualization of the structured representation to verify the correctness of the LLM's output while the law-based rules return reliable decisions. Keywords: Neuro-symbolic artificial intelligence, Knowledge graphs, Decision support systems, Tax management, Value-added tax.",

author = "Marina Luketina and Lukas Knogler and Sch{\"u}tz, \{Christoph Georg\}",

note = "Schu25a",

year = "2025",

month = dec,

language = "English",

series = "CEUR Workshop Proceedings",

publisher = "CEUR Workshop Proceedings (CEUR-WS.org)",

pages = "130--142",

editor = "\{Novakazi \}, \{Fjoll{\"e} \} and \{Dalal \}, \{Aryan Singh\}",

booktitle = "Joint Proceedings of the 16th Workshop on Ontology Design and Patterns and the 1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025), Nara, Japan, November 2-3, 2025",

edition = "1",

}

Luketina, M, Knogler, L & Schütz, CG 2025, Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law. in F Novakazi & AS Dalal (eds), Joint Proceedings of the 16th Workshop on Ontology Design and Patterns and the 1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025), Nara, Japan, November 2-3, 2025. 1 edn, CEUR Workshop Proceedings, vol. 4093, CEUR Workshop Proceedings (CEUR-WS.org), pp. 130-142. <https://ceur-ws.org/Vol-4093/Paper5hai.pdf>

Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law. / Luketina, Marina; Knogler, Lukas; Schütz, Christoph Georg.
Joint Proceedings of the 16th Workshop on Ontology Design and Patterns and the 1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025), Nara, Japan, November 2-3, 2025. ed. / Fjollë Novakazi ; Aryan Singh Dalal . 1. ed. CEUR Workshop Proceedings (CEUR-WS.org), 2025. p. 130-142 (CEUR Workshop Proceedings; Vol. 4093).

Research output: Chapter in Book/Report/Conference proceeding › Conference proceedings › peer-review

TY - GEN

T1 - Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law

AU - Luketina, Marina

AU - Knogler, Lukas

AU - Schütz, Christoph Georg

N1 - Schu25a

PY - 2025/12

Y1 - 2025/12

N2 - In tax advisory practice, case descriptions are typically not structured in a machine-readable format, with clients describing their situation in natural language. Large language models excel at natural-language understanding. However, for legal reasoning, including tax law, the propensity of LLMs to hallucinate presents a considerable challenge. Rule-based systems, on the other hand, offer verifiably correct reasoning given the correct input. Therefore, in this paper, we propose a hybrid approach to support tax advisors with analyzing tax cases, combining a rule-based system with large language models. We focus on the analysis of chain-transaction cases in value-added tax (VAT) law, where the law states a clear set of rules for regular chain-transaction cases. We employ a large language model (LLM) for the construction of structured representations of natural-language VAT case descriptions and law-based rules for the identification of the movable supply, which determines tax liabilities. Human tax advisors can obtain a graphical visualization of the structured representation to verify the correctness of the LLM's output while the law-based rules return reliable decisions. Keywords: Neuro-symbolic artificial intelligence, Knowledge graphs, Decision support systems, Tax management, Value-added tax.

AB - In tax advisory practice, case descriptions are typically not structured in a machine-readable format, with clients describing their situation in natural language. Large language models excel at natural-language understanding. However, for legal reasoning, including tax law, the propensity of LLMs to hallucinate presents a considerable challenge. Rule-based systems, on the other hand, offer verifiably correct reasoning given the correct input. Therefore, in this paper, we propose a hybrid approach to support tax advisors with analyzing tax cases, combining a rule-based system with large language models. We focus on the analysis of chain-transaction cases in value-added tax (VAT) law, where the law states a clear set of rules for regular chain-transaction cases. We employ a large language model (LLM) for the construction of structured representations of natural-language VAT case descriptions and law-based rules for the identification of the movable supply, which determines tax liabilities. Human tax advisors can obtain a graphical visualization of the structured representation to verify the correctness of the LLM's output while the law-based rules return reliable decisions. Keywords: Neuro-symbolic artificial intelligence, Knowledge graphs, Decision support systems, Tax management, Value-added tax.

M3 - Conference proceedings

T3 - CEUR Workshop Proceedings

SP - 130

EP - 142

BT - Joint Proceedings of the 16th Workshop on Ontology Design and Patterns and the 1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025), Nara, Japan, November 2-3, 2025

A2 - Novakazi , Fjollë

A2 - Dalal , Aryan Singh

PB - CEUR Workshop Proceedings (CEUR-WS.org)

ER -

Luketina M, Knogler L, Schütz CG. Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law. In Novakazi F, Dalal AS, editors, Joint Proceedings of the 16th Workshop on Ontology Design and Patterns and the 1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025), Nara, Japan, November 2-3, 2025. 1 ed. CEUR Workshop Proceedings (CEUR-WS.org). 2025. p. 130-142. (CEUR Workshop Proceedings).

Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law

Abstract

Publication series

Fields of science

JKU Focus areas

UN SDGs

Access to Document

Activities

Using Large Language Models and Law-Based Rules for the Analysis of VAT Chain-Transaction Cases in Austrian Tax Law

1st Workshop on Bridging Hybrid Intelligence and the Semantic Web (WOP-HAIBRIDGE 2025) co-located with the 24th International Semantic Web Conference (ISWC 2025)

Cite this