Create Readme file
Browse files
README.md
ADDED
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Basic Information
|
2 |
+
|
3 |
+
This is the Dr. Decr-large model used in XOR-TyDi leaderboard task 1 whitebox submission.
|
4 |
+
|
5 |
+
https://nlp.cs.washington.edu/xorqa/
|
6 |
+
|
7 |
+
|
8 |
+
The detailed implementation of the model can be found in:
|
9 |
+
|
10 |
+
https://arxiv.org/pdf/2112.08185.pdf
|
11 |
+
|
12 |
+
Source code to train the model can be found via PrimeQA's IR component:
|
13 |
+
https://github.com/primeqa/primeqa/tree/main/examples/drdecr
|
14 |
+
|
15 |
+
It is a Neural IR model built on top of the ColBERTv2 api and not directly compatible with Huggingface API. The inference result on XOR Dev dataset is:
|
16 |
+
```
|
17 |
+
R@2kt R@5kt
|
18 |
+
ko 69.1 75.1
|
19 |
+
ar 68.0 75.7
|
20 |
+
bn 81.9 85.2
|
21 |
+
fi 68.2 73.6
|
22 |
+
ru 67.1 72.2
|
23 |
+
ja 63.1 69.7
|
24 |
+
te 82.8 86.1
|
25 |
+
Avg 71.4 76.8
|
26 |
+
```
|
27 |
+
# Limitations and Bias
|
28 |
+
|
29 |
+
This model used pre-trained XLMR-large model and fine tuned on 7 languages in XOR-TyDi leaderboard. The performance of other languages was not tested.
|
30 |
+
|
31 |
+
Since the model was fine-tuned on a large pre-trained language model XLM-Roberta, biases associated with the pre-existing XLM-Roberta model may be present in our fine-tuned model, Dr. Decr
|
32 |
+
|
33 |
+
# Citation
|
34 |
+
```
|
35 |
+
@article{Li2021_DrDecr,
|
36 |
+
doi = {10.48550/ARXIV.2112.08185},
|
37 |
+
url = {https://arxiv.org/abs/2112.08185},
|
38 |
+
author = {Li, Yulong and Franz, Martin and Sultan, Md Arafat and Iyer, Bhavani and Lee, Young-Suk and Sil, Avirup},
|
39 |
+
keywords = {Computation and Language (cs.CL), Artificial Intelligence (cs.AI), FOS: Computer and information sciences, FOS: Computer and information sciences},
|
40 |
+
title = {Learning Cross-Lingual IR from an English Retriever},
|
41 |
+
publisher = {arXiv},
|
42 |
+
year = {2021}
|
43 |
+
}
|
44 |
+
```
|