Identifying rhetorical roles in legal judgments is a foundational step for automating legal reasoning, summarization, and retrieval. In this paper, we propose a novel Sequential Tolerance-based Classifier (SeqTNS) for rhetorical role classification in Indian legal documents. The proposed classifier leverages semantic similarity and contextual dependencies by using label sequence aware BiLSTMs on top of word embeddings from finetuned InLegalBERT model. These enriched embeddings are clustered into tolerance classes via a tolerance relation using a cosine distance threshold,enabling the model to make flexible, similarity-based predictions. We evaluate SeqTNS on two benchmark datasets annotated with thirteen and seven rhetorical roles, respectively. The proposed method outperforms fine-tuned transformer baselines (LegalBERT, InLegalBERT) as well as the previously developed tolerance relation-based (TNS) model, achieving a weighted F1 score of 0.78 on thirteen class dataset and a macro F1 of 0.83 on the seven class dataset, while reducing training time by 39-40% compared to state of the art BiLSTM-CRF models. The larger of our two datasets is substantial, containing over 40,000 sentences and 1.3M tokens, and serves as a challenging real world benchmark. Additionally, we use LIME for explainability and t-SNE to validate the coherence of tolerance-based clusters.

This paper presents the systems we submitted to the JUST-NLP 2025 Shared Task on Legal Summarization (L-SUMM). Creating abstractive summaries of lengthy Indian court rulings is challenging due to transformer token limits. To address this problem, we compare three systems built on a fine-tuned Legal Pegasus model. System 1 (Baseline) applies a standard hierarchical framework that chunks long documents using naive token-based segmentation. System 2 (RR-Chunk) improves this approach by using a BERT-BiLSTM model to tag sentences with rhetorical roles (RR) and incorporating these tags (e.g., [Facts]. . . ) to enable structurally informed chunking for hierarchical summarization. System 3 (WRR-Tune) tests whether explicit importance cues help the model by assigning importance scores to each RR using the geometric mean of their distributional presence in judgments and human summaries, and finetuning a separate model on text augmented with these tags (e.g., [Facts, importance score 13.58]). A comparison of the three systems demonstrates the value of progressively adding structural and quantitative importance signals to the model’s input.