Extractive question answering benchmark covers all indian intellectual property laws
IndicIPR-QA: A Benchmark for Extractive Question Answering on Indian IPR Laws
Computation and Language
Summary
Understanding legal questions about India’s intellectual property rights laws is tough because the laws are long and complicated. The authors created a new dataset with thousands of question-and-answer pairs drawn directly from the exact legal texts of seven important Indian IPR laws. Their method uses AI to generate and check questions and answers, ensuring high quality. They tested different ways to find answers, showing that simple matching doesn’t work well and more advanced AI models do much better. This work helps build smarter tools to find precise legal answers quickly in complex Indian laws.
What this means in practice
- •For legal technology developers: Build tools that extract precise answers from Indian intellectual property laws to assist lawyers in legal research and compliance checking.
- •For government legal departments: Automate answering detailed legal questions about Indian intellectual property statutes to improve public legal services and internal knowledge management.
Authors
M. Mohaiminul Islam, Muhammad Abulaish
Abstract
Legal Question Answering (QA) over statutory texts is a challenging task in Natural Language Processing (NLP), especially in low-resource settings with complex legal systems. It requires answers that are precise and grounded in the law. However, to the best of current knowledge, no existing extractive QA benchmark provides exact answer spans from statutory texts across all seven Indian Intellectual Property Rights (IPR) Acts. This paper presents IndicIPR-QA1, an extractive QA benchmark for Indian IPR laws. The proposed dataset contains 4,462 QA pairs from 1,116 statutory contexts across seven Acts. Questions and answers are generated using GPT-4o and validated through nine automatic quality checks, single and multi-model LLM evaluations, and a 200-pair human audit. Section-level partitioning prevents source-context overlap across the training, validation, and test sets. Further analysis examines answer span length, lexical overlap, and multi-clause structure across the corpus and individual Acts. BM25 achieves a token-level F1 score of 20.21, while fine-tuned InLegalBERT reaches 85.02 under a unified sliding-window inference protocol. The large performance gap between BM25 (20.21 F1) and InLegalBERT (85.02 F1) shows that simple word matching is not enough to answer these legal questions. Performance also varies across the seven Acts, indicating substantial differences in answer-extraction performance.