notesum.ai
Published at December 10Bilingual BSARD: Extending Statutory Article Retrieval to Dutch
cs.CL
cs.IR
Released Date: December 10, 2024
Authors: Ehsan Lotfi1, Nikolay Banar1, Nerses Yuzbashyan1, Walter Daelemans1
Aff.: 1University of Antwerp

| T | Model | Size | R@100 | R@200 | R@500 | MAP@100 | MRR@100 | nDCG@10 | nDCG@100 |
|---|---|---|---|---|---|---|---|---|---|
| TF-IDF | - | 39.21 | 46.38 | 52.76 | 8.53 | 14.25 | 12.38 | 16.74 | |
| BM25 | - | 40.19 | 47.95 | 54.57 | 16.07 | 22.63 | 20.07 | 23.57 | |
| word2vec | - | 41.06 | 51.02 | 58.94 | 8.28 | 15.27 | 11.66 | 17.05 | |
| fastText | - | 31.47 | 38.26 | 49.08 | 7.27 | 12.67 | 8.89 | 13.86 | |
| mE5small | 118M | 45.43 | 52.25 | 61.10 | 13.42 | 21.79 | 17.67 | 22.79 | |
| mContriever | 178M | 47.92 | 58.38 | 68.32 | 11.38 | 20.15 | 14.83 | 21.82 | |
| DPR-XM | 277M | 40.44 | 46.12 | 53.16 | 13.57 | 21.78 | 16.40 | 21.79 | |
| mE5base | 278M | 50.14 | 57.68 | 65.30 | 16.47 | 25.64 | 20.81 | 26.49 | |
| mGTE | 305M | 52.78 | 61.97 | 73.09 | 15.86 | 24.92 | 20.08 | 26.80 | |
| LaBSE | 471M | 20.51 | 28.42 | 42.18 | 2.34 | 6.60 | 3.50 | 7.18 | |
| mE5large | 560M | 58.35 | 65.83 | 70.83 | 21.88 | 34.28 | 28.47 | 33.51 | |
| mE5large-instruct | 560M | 59.48 | 66.80 | 75.21 | 18.66 | 29.93 | 24.84 | 31.33 | |
| BGE-M3 | 568M | 61.12 | 67.20 | 77.56 | 18.31 | 30.40 | 24.04 | 31.21 | |
| jina-embeddings-v3 | 572M | 60.70 | 67.92 | 77.37 | 18.59 | 31.21 | 24.70 | 31.58 | |
| E5mistral-7b | 7B | 68.35 | 73.91 | 82.82 | 30.24 | 43.26 | 37.70 | 43.02 | |
| BGE-Mult.-Gemma2 | 9B | 69.94 | 76.23 | 81.28 | 25.07 | 37.66 | 30.95 | 39.11 | |
| voyage-3 | - | 73.08 | 79.37 | 85.67 | 32.81 | 46.38 | 40.06 | 46.21 | |
| embedding-3-large | - | 75.70 | 80.22 | 88.24 | 29.73 | 42.99 | 36.83 | 44.40 | |
| ✓ | Tik-to-Tokbase | 116M | 73.90 | 79.02 | 83.29 | 39.24 | 45.69 | 42.75 | 49.90 |
| ✓ | RobBERT-2023base | 125M | 75.08 | 79.33 | 83.40 | 40.51 | 47.68 | 44.76 | 51.36 |
| ✓ | XLM-Robertabase | 279M | 62.06 | 68.26 | 75.40 | 26.61 | 32.00 | 30.65 | 37.10 |