Update README.md
Browse files
README.md
CHANGED
@@ -1,10 +1,17 @@
|
|
1 |
-
# EC40
|
|
|
|
|
2 |
|
3 |
### EC40 is an English-Centric Multilingual Machine Translation Dataset. It has over 60 Million sentences including 40 Languages across 5 Language Families.
|
4 |
|
5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
|
7 |
-
<br>
|
8 |
|
9 |
-----
|
10 |
## Languages and Family
|
@@ -15,10 +22,6 @@
|
|
15 |
| Romance | French, Spanish, Italian, Portuguese, Romanian, Occitan, Asturian, Catalan |
|
16 |
| Slavic | Russian, Czech, Polish, Bulgarian, Ukrainian, Serbian, Belarusian, Bosnian |
|
17 |
| Indo-Aryan | Hindi, Bengali, Kannada, Marathi, Sindhi, Gujarati, Nepali, Urdu |
|
18 |
-
|
|
19 |
-
|
20 |
-
<br>
|
21 |
-
|
22 |
-----
|
23 |
|
24 |
## Dataset Stats
|
@@ -30,10 +33,8 @@
|
|
30 |
| Low | af, lb, ro, oc, uk, sr, sd, gu, ti, am | 100k |
|
31 |
| Extremely-Low | no, is, ast, ca, be, bs, ne, ur, kab, so | 50k |
|
32 |
|
33 |
-
<br>
|
34 |
-
|
35 |
-----
|
36 |
-
## Build Fairseq dataset (
|
37 |
|
38 |
```
|
39 |
Read toolkit/build_fairseq_sharded_dataset.sh
|
@@ -46,4 +47,4 @@ Read toolkit/build_fairseq_sharded_dataset.sh
|
|
46 |
|
47 |
```
|
48 |
Read toolkit/train-EC40-mTrans-large.sh
|
49 |
-
```
|
|
|
1 |
+
# EC40 MNMT Dataset
|
2 |
+
|
3 |
+
GitHub: https://github.com/Smu-Tan/ZS-NMT-Variations/tree/main
|
4 |
|
5 |
### EC40 is an English-Centric Multilingual Machine Translation Dataset. It has over 60 Million sentences including 40 Languages across 5 Language Families.
|
6 |
|
7 |
+
#### Note: The dataset is cleaned and pre-processed using tools like Moses, for more details, please refer to the paper.
|
8 |
+
|
9 |
+
### Features:
|
10 |
+
1. We carefully balanced the dataset across resources and languages by strictly maintaining each resource group containing 5 language families and each family consists of 8 representative languages.
|
11 |
+
2. EC40 covers a wide spectrum of resource availability, ranging from High(5M) to Medium(1M), Low(100K), and extremely-Low(50K) resources.
|
12 |
+
3. In total, there are 80 English-centric directions for training and 1,640 directions (including all supervised and ZS directions) for evaluation.
|
13 |
+
4. We make use of Ntrex-128 and Flores-200 as our validation and test set.
|
14 |
|
|
|
15 |
|
16 |
-----
|
17 |
## Languages and Family
|
|
|
22 |
| Romance | French, Spanish, Italian, Portuguese, Romanian, Occitan, Asturian, Catalan |
|
23 |
| Slavic | Russian, Czech, Polish, Bulgarian, Ukrainian, Serbian, Belarusian, Bosnian |
|
24 |
| Indo-Aryan | Hindi, Bengali, Kannada, Marathi, Sindhi, Gujarati, Nepali, Urdu |
|
|
|
|
|
|
|
|
|
25 |
-----
|
26 |
|
27 |
## Dataset Stats
|
|
|
33 |
| Low | af, lb, ro, oc, uk, sr, sd, gu, ti, am | 100k |
|
34 |
| Extremely-Low | no, is, ast, ca, be, bs, ne, ur, kab, so | 50k |
|
35 |
|
|
|
|
|
36 |
-----
|
37 |
+
## Build Fairseq dataset (Shard->to avoid RAM OOM)
|
38 |
|
39 |
```
|
40 |
Read toolkit/build_fairseq_sharded_dataset.sh
|
|
|
47 |
|
48 |
```
|
49 |
Read toolkit/train-EC40-mTrans-large.sh
|
50 |
+
```
|