Explainable Machine Learning Multi-Label Classification of Spanish Legal Judgements - Hybrid System for Natural-Language Explanations

Aug 4, 2026

Upload by Hazel

Artificial Intelligence techniques such as Machine Learning (ML) are not fully exploited in the legal domain due to limited decision explanations. Automatic expert systems with explanatory capabilities can support legal professionals when searching jurisprudence for contextual knowledge. The work proposes a hybrid approach using ML for multi-label classification of judgements (sentences) and visual and natural language descriptions, powered by NLP and deep legal reasoning to identify involved entities. Experiments reach over 85% micro precision on expert-annotated data, reducing labor-intensive classification work.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Legal data analytics and limitations of human identification

Natural language processing and machine learning in legal tasks

Multi-label classification for legal judgements

Rationale: multiple law categories per judgement

Challenges: imbalance and feature engineering

Proposed explainable hybrid system

Entity detection for deep legal reasoning

Explainable machine learning multi-label classiﬁcation of Spanish legal judgements Francisco de Arriba-Pérez 1, Silvia García-Méndez ⇑, 1, Francisco J. González-Castaño 1, Jaime González-González 1 Information Technologies Group, atlanTTic, University of Vigo, EI Telecomunicación, Campus Lagoas-Marcosende, Vigo 36310, Spain | a r t i c l e i n f o | | a b s t r a c t | | --- | --- | --- | | Article history:
Received 11 August 2022
Revised 23 September 2022
Accepted 17 October 2022
Available online 27 October 2022 | | Artiﬁcial Intelligence techniques such as Machine Learning (ML) have not been exploited to their maximum potential in the legal domain. This has been partially due to the insufﬁcient explanations they provided about their decisions. Automatic expert systems with explanatory capabilities can be specially useful when legal practitioners search jurisprudence to gather contextual knowledge for their cases. Therefore, we propose a hybrid system that applies ML for multi-label classiﬁcation of judgements (sentences) and visual and natural language descriptions for explanation purposes, boosted by Natural Language Processing techniques and deep legal reasoning to identify the entities, such as the parties, involved. We are not aware of any prior work on automatic multi-label classiﬁcation of legal judgements also providing natural language explanations to the end-users with comparable overall quality. Our solution achieves over 85% micro precision on a labelled data set annotated by legal experts. This endorses its interest to relieve human experts from monotonous labour-intensive legal classiﬁcation tasks.
􀀁 2022 The Author(s). Published by Elsevier B.V. on behalf of King Saud University. This is an open access article under the CC BY-NC-ND license ([http://creativecommons.org/licenses/by-nc-nd/4.0/](http://creativecommons.org/licenses/by-nc-nd/4.0/)). | | Keyword:
Machine learning
Natural language processing Multi-label text classiﬁcation Interpretability and explainability Legal texts | | | 1. Introduction Artiﬁcial Intelligence (AI) techniques have not been exploited to their maximum potential in the legal ﬁeld (Qiu et al., 2020) despite the vast amounts of judgements (sentences) that lawyers must review (Zhou et al., 2022). For example, between 2013 and 2016, the number of cases in the Spanish legal system increased by 141%2. Thus, automatic legal data analytics deserve attention to improve its productiveness. This type of analysis must rely on existing legal databases (Csányi et al., 2022), which organise judgements by jurisdictions. These databases are helpful as contextual information when legal practitioners such as lawyers, barristers and solicitors look for similar court decisions. However, human-conducted identiﬁcation tends to be biased by personal points of view and this is exacerbated by the different styles of legal writing and lack of a uniﬁed formal structure (Csányi et al., 2022). Nowadays, many industrial sectors take advantage of the mostrecent advances in Natural Language Processing (NLP) techniques ⇑ Corresponding author. E-mail address: [sgarcia@gti.uvigo.es](sgarcia@gti.uvigo.es) (S. García-Méndez). 1 These authors contributed equally to this work. 2 Available at [https://www.poderjudicial.es/cgpj/es/Poder-Judicial/Tribunal](https://www.poderjudicial.es/cgpj/es/Poder-Judicial/Tribunal)Supremo/Portal-de-Transparencia/Te-puede-interesar—/Estadisticas-, August 2022. and Machine Learning (ML) algorithms (Oussous et al., 2018; Tanget al., 2020; Roh et al., 2021), and the legal ﬁeld is no exception, for example for summarisation (Kanapala et al., 2019) and anonymisation (Di Martino et al., 2021) purposes. Previous work has already addressed multi-label legal document classiﬁcation (Song et al., 2022). However, there are few freely accessible annotated data sets, their quality is often low, and they tend to be either over-or under-inclusive. This is the c

FAQs

Why do machine learning methods need explainability in the legal domain?

Legal practitioners require understandable reasons for decisions when using models. Limited explanations reduce trust and practical usefulness in jurisprudence exploration and legal case support.

What problem does the proposed hybrid system address?

It performs multi-label classification of legal judgements and generates natural-language and visual explanations. The method uses NLP and deep legal reasoning to identify relevant entities such as the parties involved.

How is the approach evaluated and what performance is reported?

Evaluation uses a labeled dataset annotated by legal experts. The system achieves over 85% micro precision, demonstrating strong overall quality and potential to reduce manual, labor-intensive classification work.

Aug 4, 2026

Upload by Hazel