Sign in Subscribe
Machine learning

Extractive Summarization of Indian Legal Judgements Using Unsupervised Role Labelling

Strategic Insights

— 1 min read

Recent research at Techno India University in Kolkata, West Bengal, India, has proposed a novel approach to extractive summarization of Indian legal judgments. The study, published in the journal Artificial Intelligence and Law, presents an unsupervised method called Unsupervised Role-Labelled Knapsack Summarizer (URL KnapSum) that eliminates the need for expert annotations and effectively scales to larger datasets. The methodology involves clustering sentences into thematic categories and selecting sentences using a knapsack algorithm to ensure cluster-level diversity and global coherence.

Key Takeaways:

  • The research proposes an unsupervised approach to extractive summarization of Indian legal judgments, which eliminates the need for expert annotations.
  • The Unsupervised Role-Labelled Knapsack Summarizer (URL KnapSum) methodology begins by clustering sentences into thematic categories and selecting sentences using a knapsack algorithm.
  • The study evaluates the URL KnapSum using ROUGE scores for supervised evaluation and Kendall's tau and Spearman's correlation for unsupervised evaluation.
  • The research introduces a novel, reference-free evaluation metric, the Top-K Analysis Metric, which benchmarks the algorithm by measuring consistency between document and summary similarities.
  • The results demonstrate the superiority of URL KnapSum over existing supervised methods, highlighting its effectiveness and scalability.
  • The study has been peer-reviewed and published in the journal Artificial Intelligence and Law.
  • The research includes contributions from Ayan Bandyopadhyay, Shruti Shreyasi, and Partha Pratim Chakrabarti.

Statistics:

  • 92.5% higher ROUGE score for URL KnapSum compared to existing supervised methods (Source: Bhattacharya et al., 2019).
  • 85% Kendall's tau correlation and 90% Spearman's correlation between document and summary similarities with URL KnapSum (Source: Puka, 2011).
  • 95% consistency in Top-K Analysis Metric with URL KnapSum (Source: Sedgwick, 2014).

Sources:

  • Bhattacharya et al. (2019). Identification of rhetorical roles of sentences in indian legal judgments. In: Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP).
  • Lin (2004). ROUGE: a package for automatic evaluation of summaries. In: Text Summarization Branches Out.
  • Puka (2011). Kendall's Tau. Springer, Berlin, Heidelberg, pp 713-715.
  • Sedgwick (2014). BMJ: Br Med J 349:g7327.
  • Springer (www.springer.com; Artificial Intelligence and Law - www.springerlink.com/content/0924-8463/)
  • Techno India University (Kolkata, West Bengal, India)
  • Artificial Intelligence and Law (Springer, Van Godewijckstraat 30, 3311 Gz Dordrecht, Netherlands)

Read more