3 papers accepted @ COLING 2025! Click for more!
๐ 3 papers accepted @ COLING 2025!
Just before the holiday break๐โ๏ธ and I am delighted to share that some of our latest NLP work is making waves ๐ since we got 3(!) papers accepted at the 31st International Conference on Computational Linguistics (COLING 2025 https://lnkd.in/dR8bgtXu)! Some info on the exceptional work of these A-๐-M-๐ซ-A-โจ-Z-๐ฅ-I-๐-N-๐-G-๐ PhD candidates below ๐๐๐
1/ The ultimate PhD paper of Antoine Louis asks โwhether to fuse or not to fuseโ in an (legal) IR scenario ๐. BM25 is still a performance beast ๐ช in IR, but itโs crucial to know when it shines โจ and where it falls short โ ๏ธ compared to dense models. In the paper we explore different scenarios and conclude: โ BM25 = still the ๐ of search, esp. in zero-shot tasks or when efficiency rules. โ Fusing models? ๐ค Great for zero-shotโboosts general IR models โ Got domain-specific data? ๐ง Fine-tune one model forbest results.
๐ Paper: https://lnkd.in/dxr2VQQE (w/Gijs van Dijck) ๐ป Code: https://lnkd.in/d8wrEP7i ๐ค Models: https://lnkd.in/d4RwVVfc
2/ Have you noticed how most information retrieval work is in English and Chinese? Well, Antoine and Vageesh noticed the same and as a PhD-side-project worked on delivering ColBERT-XM ๐ a modular retriever for 81+ languages ๐งฉ Built w/XMOD encoders & ColBERTโs backbone, it trains on English (high-resource language) and transfers zero-shot to other languages, thereby eliminating the need for language-specific labeled retrieval data.๐กโจ
๐ Paper: https://lnkd.in/dXHKunum (w/Gijs van Dijck) ๐ป Code: https://lnkd.in/dw4N5PdP ๐ค Model: https://lnkd.in/dHpAN5yR
3/ Paweล Mฤ kaโs 2nd PhD paper dives into: how do context-aware machine translation models really use context? ๐ค We analyzed attention heads and found: ๐ Some are critical for pronoun disambiguation. ๐ Fine-tuning these heads = boosted performance! This work builds on VOXReality EU project, where we -efficiently- integrate SoTA MT models in AR/VR ๐ถ๏ธ๐๐ฎ scenarios, therefore context use is essential.
๐ Paper: https://lnkd.in/dc9sVYtn (w/Yusuf Can Semerci, Johannes (Jan) C. Scholtes) ๐ป Code: https://lnkd.in/duNdb5YY