@inproceedings{torres-etal-2026-hybrid,
title = "A Hybrid Framework for Precedent Reference Extraction",
author = "Torres, Arthur Elwing and
Moreira, Johny and
Vianna, Daniela and
Masid, Marina Ramalhete and
Larcher, Celio and
de Moura, Edleno Silva and
Silva, Altigran da",
editor = "Atanasova, Pepa and
Haffari, Gholamreza and
Wang, Minghan",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-industry.152/",
pages = "2582--2590",
ISBN = "979-8-89176-460-6",
abstract = "Legal documents frequently contain heterogeneous references to judicial precedents, making automated Precedent Reference Extraction (PRE) challenging for legal AI systems. Existing approaches rely heavily on LLM-based pipelines, which are often effective but also incur recurring prompting costs in high-volume production workloads and struggle with tracing extracted entities back to their character offsets in the source text. In light of these issues, we present $\textit{Notarius}$, a domain-specific and production-ready hybrid framework that decomposes PRE into two stages: Overlapped NER Feature Extraction (ONFE), which extracts fine-grained precedent features, and Precedent Feature Aggregation (PFA), which reconstructs complete references for downstream applications. Our implementation integrates the Unified NER method W$^2$NER with domain-specific heuristics. We evaluate $\textit{Notarius}$ at both feature and reference levels on a collection of judicial decisions and compare it in production against an LLM-based pipeline. $\textit{Notarius}$ achieves higher extraction quality, eliminates recurring LLM API costs for PRE, and provides span-level offset traceability. The system currently powers precedent extraction on Jusbrasil, demonstrating that discriminative, domain-specialized NER architectures remain valuable for cost-effective, high-precision legal information extraction in the LLM era."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="torres-etal-2026-hybrid">
<titleInfo>
<title>A Hybrid Framework for Precedent Reference Extraction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Arthur</namePart>
<namePart type="given">Elwing</namePart>
<namePart type="family">Torres</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Johny</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Daniela</namePart>
<namePart type="family">Vianna</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marina</namePart>
<namePart type="given">Ramalhete</namePart>
<namePart type="family">Masid</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Celio</namePart>
<namePart type="family">Larcher</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Edleno</namePart>
<namePart type="given">Silva</namePart>
<namePart type="family">de Moura</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Altigran</namePart>
<namePart type="given">da</namePart>
<namePart type="family">Silva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pepa</namePart>
<namePart type="family">Atanasova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gholamreza</namePart>
<namePart type="family">Haffari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minghan</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-460-6</identifier>
</relatedItem>
<abstract>Legal documents frequently contain heterogeneous references to judicial precedents, making automated Precedent Reference Extraction (PRE) challenging for legal AI systems. Existing approaches rely heavily on LLM-based pipelines, which are often effective but also incur recurring prompting costs in high-volume production workloads and struggle with tracing extracted entities back to their character offsets in the source text. In light of these issues, we present Notarius, a domain-specific and production-ready hybrid framework that decomposes PRE into two stages: Overlapped NER Feature Extraction (ONFE), which extracts fine-grained precedent features, and Precedent Feature Aggregation (PFA), which reconstructs complete references for downstream applications. Our implementation integrates the Unified NER method W²NER with domain-specific heuristics. We evaluate Notarius at both feature and reference levels on a collection of judicial decisions and compare it in production against an LLM-based pipeline. Notarius achieves higher extraction quality, eliminates recurring LLM API costs for PRE, and provides span-level offset traceability. The system currently powers precedent extraction on Jusbrasil, demonstrating that discriminative, domain-specialized NER architectures remain valuable for cost-effective, high-precision legal information extraction in the LLM era.</abstract>
<identifier type="citekey">torres-etal-2026-hybrid</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-industry.152/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>2582</start>
<end>2590</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T A Hybrid Framework for Precedent Reference Extraction
%A Torres, Arthur Elwing
%A Moreira, Johny
%A Vianna, Daniela
%A Masid, Marina Ramalhete
%A Larcher, Celio
%A de Moura, Edleno Silva
%A Silva, Altigran da
%Y Atanasova, Pepa
%Y Haffari, Gholamreza
%Y Wang, Minghan
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-460-6
%F torres-etal-2026-hybrid
%X Legal documents frequently contain heterogeneous references to judicial precedents, making automated Precedent Reference Extraction (PRE) challenging for legal AI systems. Existing approaches rely heavily on LLM-based pipelines, which are often effective but also incur recurring prompting costs in high-volume production workloads and struggle with tracing extracted entities back to their character offsets in the source text. In light of these issues, we present Notarius, a domain-specific and production-ready hybrid framework that decomposes PRE into two stages: Overlapped NER Feature Extraction (ONFE), which extracts fine-grained precedent features, and Precedent Feature Aggregation (PFA), which reconstructs complete references for downstream applications. Our implementation integrates the Unified NER method W²NER with domain-specific heuristics. We evaluate Notarius at both feature and reference levels on a collection of judicial decisions and compare it in production against an LLM-based pipeline. Notarius achieves higher extraction quality, eliminates recurring LLM API costs for PRE, and provides span-level offset traceability. The system currently powers precedent extraction on Jusbrasil, demonstrating that discriminative, domain-specialized NER architectures remain valuable for cost-effective, high-precision legal information extraction in the LLM era.
%U https://aclanthology.org/2026.emnlp-industry.152/
%P 2582-2590
Markdown (Informal)
[A Hybrid Framework for Precedent Reference Extraction](https://aclanthology.org/2026.emnlp-industry.152/) (Torres et al., EMNLP 2026)
ACL
- Arthur Elwing Torres, Johny Moreira, Daniela Vianna, Marina Ramalhete Masid, Celio Larcher, Edleno Silva de Moura, and Altigran da Silva. 2026. A Hybrid Framework for Precedent Reference Extraction. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 2582–2590, Budapest, Hungary. Association for Computational Linguistics.