Run
| RanxHub ID: | amdbfpse/psy/test/ikr3/bm25-bienc-mean |
|---|---|
| Version: | 1.0 |
| Description: | This run was computed using BiEnc and Mean to re-rank the top 1000 BM25 results. Original and re-ranking scores were fused using the Weighted Sum fusion algorithm provided by ranx. |
| Tags: | Fine-Tuned · Re-Ranking |
| Date: | 12 August 2022 |
| Run Authors: | Elias Bassani |
| From Paper: | A Multi-Domain Benchmark for Personalized Search Evaluation |
| Paper Authors: | Elias Bassani · Pranav Kasela · Alessandro Raganato · Gabriella Pasi |
Model
| Name: | BiEnc / Mean |
|---|---|
| Description: | BiEnc is a bi-encoder retrieval model. Mean is a personalization model that compute the users' representations as the average of the representations of their associated documents. Texts representations were computed with all-MiniLM-L6-v2. |
| Tags: | Personalization · Long-Term Personalization · Neural Network · Deep Learning · Transformer |
| Paper: | A Multi-Domain Benchmark for Personalized Search Evaluation |
| Authors: | Elias Bassani · Pranav Kasela · Alessandro Raganato · Gabriella Pasi |
Results
| MAP@100 | MRR@10 | NDCG@10 |
|---|---|---|
| 0.21974948121187401 | 0.6521017130284457 | 0.3470874949797735 |