Hassan Shavarani
commited on
Commit
•
6b897ee
1
Parent(s):
a712666
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,62 @@
|
|
| 1 |
---
|
| 2 |
license: cc-by-4.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: cc-by-4.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
datasets:
|
| 6 |
+
- CoNLL2003/AIDA
|
| 7 |
+
- Wikipedia
|
| 8 |
+
tags:
|
| 9 |
+
- SpEL
|
| 10 |
+
- Entity Linking
|
| 11 |
+
- Structured Prediction
|
| 12 |
---
|
| 13 |
+
|
| 14 |
+
## SpEL (Structured prediction for Entity Linking)
|
| 15 |
+
SpEL model finetuned on English Wikipedia as well as the training portion of CoNLL2003/AIDA.
|
| 16 |
+
It is introduced in the paper [SPEL: Structured Prediction for Entity Linking (EMNLP 2023)](https://arxiv.org/abs/2310.14684).
|
| 17 |
+
The code and data are available in [this repository](https://github.com/shavarani/SpEL).
|
| 18 |
+
|
| 19 |
+
## Evaluation Results
|
| 20 |
+
Entity Linking evaluation results of *SpEL* compared to that of the literature over AIDA test sets:
|
| 21 |
+
|
| 22 |
+
| Approach | EL Micro-F1<br/>test-a | EL Micro-F1<br/>test-b | #params<br/>on GPU | speed<br/>sec/doc |
|
| 23 |
+
|-----------------------------------------------------------------|:----------------------:|:----------------------:|:----------------------------------------:|:-----------------:|
|
| 24 |
+
| Hoffart et al. (2011) | 72.4 | 72.8 | - | - |
|
| 25 |
+
| Kolitsas et al. (2018) | 89.4 | 82.4 | 330.7M | 0.097 |
|
| 26 |
+
| Broscheit (2019) | 86.0 | 79.3 | 495.1M | 0.613 |
|
| 27 |
+
| Peters et al. (2019) | 82.1 | 73.1 | - | - |
|
| 28 |
+
| Martins et al. (2019) | 85.2 | 81.9 | - | - |
|
| 29 |
+
| van Hulst et al. (2020) | 83.3 | 82.4 | 19.0M | 0.337 |
|
| 30 |
+
| Févry et al. (2020) | 79.7 | 76.7 | - | - |
|
| 31 |
+
| Poerner et al. (2020) | 90.8 | 85.0 | 131.1M | - |
|
| 32 |
+
| Kannan Ravi et al. (2021) | - | 83.1 | - | - |
|
| 33 |
+
| De Cao et al. (2021b) | - | 83.7 | 406.3M | 40.969 |
|
| 34 |
+
| De Cao et al. (2021a)<br/>(no mention-specific candidate set) | 61.9 | 49.4 | 124.8M | 0.268 |
|
| 35 |
+
| De Cao et al. (2021a)<br/>(using PPRforNED candidate set) | 90.1 | 85.5 | 124.8M | 0.194 |
|
| 36 |
+
| Mrini et al. (2022) | - | 85.7 | (train) 811.5M<br/>(test) 406.2M | - |
|
| 37 |
+
| Zhang et al. (2022) | - | 85.8 | 1004.3M | - |
|
| 38 |
+
| Feng et al. (2022) | - | 86.3 | 157.3M | - |
|
| 39 |
+
| <hr/> | <hr/> | <hr/> | <hr/> | <hr/> |
|
| 40 |
+
| **SpEL-base** (no mention-specific candidate set) | 91.3 | 85.5 | 128.9M | 0.084 |
|
| 41 |
+
| **SpEL-base** (KB+Yago candidate set) | 90.6 | 85.7 | 128.9M | 0.158 |
|
| 42 |
+
| **SpEL-base** (PPRforNED candidate set)<br/>(context-agnostic) | 91.7 | 86.8 | 128.9M | 0.153 |
|
| 43 |
+
| **SpEL-base** (PPRforNED candidate set)<br/>(context-aware) | 92.7 | 88.1 | 128.9M | 0.156 |
|
| 44 |
+
| **SpEL-large** (no mention-specific candidate set) | 91.6 | 85.8 | 361.1M | 0.273 |
|
| 45 |
+
| **SpEL-large** (KB+Yago candidate set) | 90.8 | 85.7 | 361.1M | 0.267 |
|
| 46 |
+
| **SpEL-large** (PPRforNED candidate set)<br/>(context-agnostic) | 92.0 | 87.3 | 361.1M | 0.268 |
|
| 47 |
+
| **SpEL-large** (PPRforNED candidate set)<br/>(context-aware) | 92.9 | 88.6 | 361.1M | 0.267 |
|
| 48 |
+
|
| 49 |
+
----
|
| 50 |
+
|
| 51 |
+
## Citation
|
| 52 |
+
If you use SpEL finetuned models or data, please cite our paper:
|
| 53 |
+
|
| 54 |
+
```
|
| 55 |
+
@inproceedings{shavarani2023spel,
|
| 56 |
+
title={Sp{EL}: Structured Prediction for Entity Linking},
|
| 57 |
+
author={Shavarani, Hassan S. and Sarkar, Anoop},
|
| 58 |
+
booktitle={The 2023 Conference on Empirical Methods in Natural Language Processing},
|
| 59 |
+
year={2023},
|
| 60 |
+
url={https://arxiv.org/abs/2310.14684}
|
| 61 |
+
}
|
| 62 |
+
```
|