Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,61 @@
|
|
1 |
-
|
|
|
|
|
2 |
|
3 |
Document (context) encoder, trained on monolingual (English queries) version of the XOR-TyDi training data, as described in the XOR-TyDi task description https://nlp.cs.washington.edu/xorqa/.
|
4 |
|
5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-sa-4.0
|
3 |
+
---
|
4 |
|
5 |
Document (context) encoder, trained on monolingual (English queries) version of the XOR-TyDi training data, as described in the XOR-TyDi task description https://nlp.cs.washington.edu/xorqa/.
|
6 |
|
7 |
+
## Usage
|
8 |
+
|
9 |
+
The model is compatible with the PrimeQA DPR engine, or the Hugging Face DPR engine as shown this example https://huggingface.co/docs/datasets/faiss_es.
|
10 |
+
The model uses the tokenizer from `facebook/dpr-question_encoder-multiset-base`.
|
11 |
+
|
12 |
+
## Performance comparison
|
13 |
+
|
14 |
+
| R@5kt | R@2kt | model |
|
15 |
+
|-------|-------|-------|
|
16 |
+
| 69.6 | 62.2 | DPR, En, XOR-TyDi paper (https://arxiv.org/pdf/2010.11856.pdf, table 13) |
|
17 |
+
| 70.22 | 64.34 | DPR, En, trained on En (human) version of XOR |
|
18 |
+
|
19 |
+
|
20 |
+
## BibTeX entry and citation info
|
21 |
+
```bibtex
|
22 |
+
@inproceedings{asai-etal-2021-xor,
|
23 |
+
title = "{XOR} {QA}: Cross-lingual Open-Retrieval Question Answering",
|
24 |
+
author = "Asai, Akari and
|
25 |
+
Kasai, Jungo and
|
26 |
+
Clark, Jonathan and
|
27 |
+
Lee, Kenton and
|
28 |
+
Choi, Eunsol and
|
29 |
+
Hajishirzi, Hannaneh",
|
30 |
+
booktitle = "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
|
31 |
+
month = jun,
|
32 |
+
year = "2021",
|
33 |
+
address = "Online",
|
34 |
+
publisher = "Association for Computational Linguistics",
|
35 |
+
url = "https://aclanthology.org/2021.naacl-main.46",
|
36 |
+
doi = "10.18653/v1/2021.naacl-main.46",
|
37 |
+
pages = "547--564",
|
38 |
+
}
|
39 |
+
```
|
40 |
+
|
41 |
+
```bibtex
|
42 |
+
@inproceedings{karpukhin-etal-2020-dense,
|
43 |
+
title = "Dense Passage Retrieval for Open-Domain Question Answering",
|
44 |
+
author = "Karpukhin, Vladimir and
|
45 |
+
Oguz, Barlas and
|
46 |
+
Min, Sewon and
|
47 |
+
Lewis, Patrick and
|
48 |
+
Wu, Ledell and
|
49 |
+
Edunov, Sergey and
|
50 |
+
Chen, Danqi and
|
51 |
+
Yih, Wen-tau",
|
52 |
+
booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)",
|
53 |
+
month = nov,
|
54 |
+
year = "2020",
|
55 |
+
address = "Online",
|
56 |
+
publisher = "Association for Computational Linguistics",
|
57 |
+
url = "https://aclanthology.org/2020.emnlp-main.550",
|
58 |
+
doi = "10.18653/v1/2020.emnlp-main.550",
|
59 |
+
pages = "6769--6781",
|
60 |
+
}
|
61 |
+
```
|