SinBrief creates abstracts from Sinhala legal texts without manual data
SinBrief: A Hybrid Framework for Abstractive Text Summarisation of Sinhala Legal Documents
Computation and Language
Summary
Summarising legal documents written in Sinhala is tough because there isn’t much annotated data available and the legal language is complicated. The authors created SinBrief, a method that combines smart word connections and neural networks to pick important sentences and rewrite summaries without needing human-written training examples. They tested five sentence scoring models within their system and found it can produce summaries that use different words than the originals but remain factually accurate. This shows a promising way to create automatic summaries for legal texts in languages with limited resources.
What this means in practice
- •For legal tech developers: Build tools to automatically create readable summaries from Sinhala legal documents where annotated data is scarce.$Commercial implications: Enables commercial legal software products targeting Sinhala-speaking users with better automated document handling.
- •For translation service providers: Improve pre-processing of Sinhala legal texts by generating abstractive summaries that can reduce human effort in translation workflows.
Authors
Minduli Lasandi, Nevidu Jayatilleke
Abstract
Legal document summarisation in low-resource languages presents significant challenges due to the scarcity of annotated data and the complexity of domain-specific terminology. This paper presents SinBrief, a hybrid abstractive summarisation framework for Sinhala legal documents that does not require human-annotated training data. The proposed framework combines domain-aware word graph construction with neural sentence scoring to generate abstractive summaries from Sinhala legal text. Five sentence scoring models are evaluated within the framework: mBert, Llama 3.1, Falcon 7B, Laser, and a continually pre-trained Llama model domain-adapted to Sinhala legal text. The framework is evaluated on a Sinhala legal corpus using reference-free metrics, including Coverage, Density, Compression Ratio, SummaC, and Self-BertScore. Experimental results demonstrate that SinBrief produces summaries with lower lexical overlap than extractive baselines while maintaining factual consistency, demonstrating the viability of hybrid, largely annotation-free abstractive summarisation for low-resource legal NLP tasks.