🔍 SIGIR 2025 Paper Spotlight #2: Reproducibility Matters!
Next up in our SIGIR 2025 contributions series is our reproducibility paper:
"A Reproducibility Study of Graph-Based Legal Case Retrieval" 🎓⚖️
By Gregor Donabauer and Udo Kruschwitz
Legal case retrieval is complex — and ensuring methods are reliable, replicable, and generalizable is key for progress in this high-stakes domain. In this study, we revisited CaseLink, a graph-based retrieval approach that connects legal cases and charges through semantic relationships and reference connections.
✅ We reproduced the original CaseLink study setup
🔁 Applied it to a new dataset
📈 Enhanced the graph data representation
🧠 Integrated an open LLM into the pipeline (no limitations of closed models)
🎯 Why it matters:
Our work doesn't just test reproducibility — it pushes the method further and contributes reusable resources to the community. Open science, for real. 😎🔬🛠️
📄 Explore the pre-print version of our study here: arxiv.org/abs/2504.08400
(incl. a link to all our implementations and experimental artifacts)
🙏 A big thank-you to the original CaseLink authors — Yanran Tang, Ruihong Qiu, Hongzhi Yin, Xue Li, and Helen Huang — for their helpful and timely communication during our reproducibility study. Their SIGIR 2024 paper laid the foundation for this work.
#Reproducibility #OpenScience
#LegalTech #LegalIR #ProfessionalSearch
#Research #InformationRetrieval #IR
#GraphIR #LLMsInIR
#SIGIR #SIGIR2025
#ResearchSuccess #ResearchPaperAccepted
#InformationScienceRegensburg #StayInformed
