HKJudge: A Legal Discourse-Annotated Corpus for Interpreting What Courts Find, How They Reason, and What They Rule

2026-06-04 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

The Hong Kong Judgment Discourse Dataset (HKJudge) is introduced as the first sentence-level expert-annotated legal discourse corpus, specifically for Hong Kong judgments. It encompasses criminal judgments from all five court hierarchy levels, containing approximately 290,000 sentences and 6.5 million tokens. A two-tier discourse schema assigns each sentence one of 26 rhetorical roles and annotates three sentencing elements (charge, imprisonment term, fine) at the span level. Ten legal linguistics annotators achieved an inter-annotator agreement of κ=0.8. The work formulates two tasks, rhetorical role classification and legal element extraction, providing benchmark evaluations across four BERT-based models, two open-source LLMs, and four commercial LLMs.

Key takeaway

For NLP Engineers or AI Scientists developing legal AI systems, this work highlights the critical role of expert-annotated discourse corpora. You should consider adapting the HKJudge dataset's two-tier annotation schema for your own legal text analysis projects. This approach can significantly improve the accuracy of models predicting legal judgment outcomes and extracting specific legal elements from complex court documents.

Key insights

Expert-annotated legal discourse corpora enable advanced AI understanding of court judgments.

Principles

• Sentence-level discourse annotation models judgment structure.
• A two-tier schema captures facts, reasoning, and rulings.

Method

A two-tier discourse schema assigns 26 rhetorical roles at the sentence level and annotates three sentencing elements (charge, imprisonment term, fine) at the span level, applied by legal linguistics experts.

In practice

• Utilize a two-tier annotation schema for legal text.
• Benchmark LLMs for legal rhetorical role classification.

Topics

• Legal NLP
• Corpus Annotation
• Discourse Analysis
• Legal Judgment Prediction
• Large Language Models
• Hong Kong Law

Code references

• xuanxixi/HKJudge

Best for: Research Scientist, NLP Engineer, AI Scientist, Legal Professional

Related on AIssential

• HKJudge: A Legal Discourse-Annotated Corpus for Interpreting What Courts Find, How They Reason, and What They Rule — HKJudge provides a unique, expert-annotated corpus and schema for Hong Kong legal discourse, benchmarking LLMs against human performance.
• Combining Semantic Embeddings and Knowledge Graphs for Identifying Decision Patterns in Brazilian Judicial Decisions — Hybrid representations combining semantic embeddings and knowledge graphs enhance judicial decision pattern identification.
• VLegal-Bench: Cognitively Grounded Benchmark for Vietnamese Legal Reasoning of Large Language Models — VLegal-Bench offers a cognitively-grounded benchmark for LLMs in Vietnamese civil law, revealing specialized models outperform general ones on complex legal reasoning.
• After trial judge cites AI-generated judgment, Supreme Court asks BCI to set up expert panel to examine issue - Bar and Bench — AI-generated legal "hallucinations" necessitate judicial accountability and sovereign AI development in India.
• Democratizing Legal Analytics: Resource-Efficient Information Extraction for Brazilian Case Law — Resource-efficient LLM fine-tuning enables scalable legal analytics in low-resource languages using legacy datasets.
• Retrieval-Based Multi-Label Legal Annotation: Extensible, Data-Efficient and Hallucination-Free — Retrieval-based legal annotation offers a data-efficient, hallucination-free alternative to generative LLMs for high-cardinality label spaces.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.