IL-TUR Indian Legal NLP Benchmark (Marathi)

IL-TUR Indian Legal NLP Benchmark (Marathi)

Comprehensive Indian legal NLP benchmark with 8 tasks in Marathi including Legal NER, Rhetorical Role Prediction, Court Judgment Prediction, Bail Prediction, Legal Statute Identification, Prior Case Retrieval, Summarization, and Legal Machine Translation. Far more structured than raw court document text.

Build a legal document search engine for Maharashtra courts that enables semantic search across case law in Marathi.

Quick Start

# Access from https://github.com/Legal-NLP-EkStep
import json
with open('legal_marathi_ner.json') as f:
    data = json.load(f)
print(f"Total documents: {len(data)}")
for doc in data[:3]:
    print(f"Text: {str(doc)[:80]}...")
Modality
text
Size
Multi-task benchmark; 8 legal NLP tasks in Marathi
License
Format
JSON
Language
mr, en
Update Frequency
static
Organization
Exploration Lab (ACL 2024)

Schema

FieldTypeDescription
textstringLegal text passage in Marathi or English
task_typestringNLP task type (NER, classification, summarization)
labelstringTask-specific annotation or label

Build With This

Create an automated case similarity finder for Maharashtra lawyers that identifies relevant precedents from legal text
Develop a legal document summarizer that generates concise case briefs from lengthy Maharashtra court judgments
Build a legal entity extractor that identifies parties, judges, dates, and legal provisions from Marathi court documents

AI Use Cases

Marathi legal judgment predictionCourt document summarizationLegal named entity recognitionBail prediction for Maharashtra courtsLegal statute identification
Last verified: 2026-03-09