Upload README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,231 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
pipeline_tag: translation
|
| 3 |
+
language:
|
| 4 |
+
- multilingual
|
| 5 |
+
- en
|
| 6 |
+
- am
|
| 7 |
+
- ar
|
| 8 |
+
- so
|
| 9 |
+
- sw
|
| 10 |
+
- pt
|
| 11 |
+
- af
|
| 12 |
+
- fr
|
| 13 |
+
- zu
|
| 14 |
+
- mg
|
| 15 |
+
- ha
|
| 16 |
+
- sn
|
| 17 |
+
- arz
|
| 18 |
+
- ny
|
| 19 |
+
- ig
|
| 20 |
+
- xh
|
| 21 |
+
- yo
|
| 22 |
+
- st
|
| 23 |
+
- rw
|
| 24 |
+
- tn
|
| 25 |
+
- ti
|
| 26 |
+
- ts
|
| 27 |
+
- om
|
| 28 |
+
- run
|
| 29 |
+
- nso
|
| 30 |
+
- ee
|
| 31 |
+
- ln
|
| 32 |
+
- tw
|
| 33 |
+
- pcm
|
| 34 |
+
- gaa
|
| 35 |
+
- loz
|
| 36 |
+
- lg
|
| 37 |
+
- guw
|
| 38 |
+
- bem
|
| 39 |
+
- efi
|
| 40 |
+
- lue
|
| 41 |
+
- lua
|
| 42 |
+
- toi
|
| 43 |
+
- ve
|
| 44 |
+
- tum
|
| 45 |
+
- tll
|
| 46 |
+
- iso
|
| 47 |
+
- kqn
|
| 48 |
+
- zne
|
| 49 |
+
- umb
|
| 50 |
+
- mos
|
| 51 |
+
- tiv
|
| 52 |
+
- lu
|
| 53 |
+
- ff
|
| 54 |
+
- kwy
|
| 55 |
+
- bci
|
| 56 |
+
- rnd
|
| 57 |
+
- luo
|
| 58 |
+
- wal
|
| 59 |
+
- ss
|
| 60 |
+
- lun
|
| 61 |
+
- wo
|
| 62 |
+
- nyk
|
| 63 |
+
- kj
|
| 64 |
+
- ki
|
| 65 |
+
- fon
|
| 66 |
+
- bm
|
| 67 |
+
- cjk
|
| 68 |
+
- din
|
| 69 |
+
- dyu
|
| 70 |
+
- kab
|
| 71 |
+
- kam
|
| 72 |
+
- kbp
|
| 73 |
+
- kr
|
| 74 |
+
- kmb
|
| 75 |
+
- kg
|
| 76 |
+
- nus
|
| 77 |
+
- sg
|
| 78 |
+
- taq
|
| 79 |
+
- tzm
|
| 80 |
+
- nqo
|
| 81 |
+
|
| 82 |
+
license: apache-2.0
|
| 83 |
+
---
|
| 84 |
+
SSA-COMET-MTL, a robust, unified automatic metric for both MTE and QE, built based on SSA-MTE: It receives a triplet with (source sentence, translation, reference translation) for MTE, or a pair with (source sentence, translation) for QE, and returns a score that reflects the quality of the translation.
|
| 85 |
+
This model is based on an improved African enhanced encoder, [afro-xlmr-large-76L](https://huggingface.co/Davlan/afro-xlmr-large-76L).
|
| 86 |
+
|
| 87 |
+
# Paper
|
| 88 |
+
|
| 89 |
+
Li S., Wang J., Ali F., Cherry C., Deutsch D., Briakou E., Sousa-Silva R., Cardoso H.L, Stenetorp P., and Adelani D.I.
|
| 90 |
+
[SSA-COMET: Do LLMs Outperform Learned Metrics in Evaluating MT for Under-Resourced African Languages?](https://aclanthology.org/2025.emnlp-main.656/). In EMNLP 2025
|
| 91 |
+
|
| 92 |
+
|
| 93 |
+
# License
|
| 94 |
+
|
| 95 |
+
Apache-2.0
|
| 96 |
+
|
| 97 |
+
# Usage (SSA-COMET)
|
| 98 |
+
|
| 99 |
+
Using this model requires unbabel-comet to be installed:
|
| 100 |
+
|
| 101 |
+
```bash
|
| 102 |
+
pip install --upgrade pip # ensures that pip is current
|
| 103 |
+
pip install unbabel-comet
|
| 104 |
+
```
|
| 105 |
+
|
| 106 |
+
Then you can use it through comet CLI:
|
| 107 |
+
|
| 108 |
+
```bash
|
| 109 |
+
comet-score -s {source-inputs}.txt -t {translation-outputs}.txt -r {references}.txt --model McGill-NLP/ssa-comet-mtl
|
| 110 |
+
```
|
| 111 |
+
|
| 112 |
+
Or using Python:
|
| 113 |
+
|
| 114 |
+
```python
|
| 115 |
+
from comet import download_model, load_from_checkpoint
|
| 116 |
+
model_path = download_model("McGill-NLP/ssa-comet-mtl")
|
| 117 |
+
model = load_from_checkpoint(model_path)
|
| 118 |
+
data = [
|
| 119 |
+
{
|
| 120 |
+
"src": "Nadal sàkọọ́lẹ̀ ìforígbárí o ní àmì méje sóódo pẹ̀lú ilẹ̀ Canada.",
|
| 121 |
+
"mt": "Nadal's head to head record against the Canadian is 7–2.",
|
| 122 |
+
"ref": "Nadal scored seven unanswered points against Canada."
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"src": "Laipe yi o padanu si Raoniki ni ere Sisi Brisbeni.",
|
| 126 |
+
"mt": "He recently lost against Raonic in the Brisbane Open.",
|
| 127 |
+
"ref": "He recently lost to Raoniki in the game Sisi Brisbeni."
|
| 128 |
+
}
|
| 129 |
+
]
|
| 130 |
+
model_output = model.predict(data, batch_size=8, gpus=1)
|
| 131 |
+
print (model_output)
|
| 132 |
+
```
|
| 133 |
+
|
| 134 |
+
# Intended uses
|
| 135 |
+
|
| 136 |
+
Our model is intended to be used for **MT evaluation** and **Quality Eestimation**.
|
| 137 |
+
|
| 138 |
+
Given a triplet with (source sentence, translation, reference translation), or a pair with (source sentence, translation), it outputs a single score between 0 and 1, where 1 represents a perfect translation.
|
| 139 |
+
|
| 140 |
+
# Languages Covered:
|
| 141 |
+
|
| 142 |
+
There are 76 languages available :
|
| 143 |
+
- English (eng)
|
| 144 |
+
- Amharic (amh)
|
| 145 |
+
- Arabic (ara)
|
| 146 |
+
- Somali (som)
|
| 147 |
+
- Kiswahili (swa)
|
| 148 |
+
- Portuguese (por)
|
| 149 |
+
- Afrikaans (afr)
|
| 150 |
+
- French (fra)
|
| 151 |
+
- isiZulu (zul)
|
| 152 |
+
- Malagasy (mlg)
|
| 153 |
+
- Hausa (hau)
|
| 154 |
+
- chiShona (sna)
|
| 155 |
+
- Egyptian Arabic (arz)
|
| 156 |
+
- Chichewa (nya)
|
| 157 |
+
- Igbo (ibo)
|
| 158 |
+
- isiXhosa (xho)
|
| 159 |
+
- Yorùbá (yor)
|
| 160 |
+
- Sesotho (sot)
|
| 161 |
+
- Kinyarwanda (kin)
|
| 162 |
+
- Tigrinya (tir)
|
| 163 |
+
- Tsonga (tso)
|
| 164 |
+
- Oromo (orm)
|
| 165 |
+
- Rundi (run)
|
| 166 |
+
- Northern Sotho (nso)
|
| 167 |
+
- Ewe (ewe)
|
| 168 |
+
- Lingala (lin)
|
| 169 |
+
- Twi (twi)
|
| 170 |
+
- Nigerian Pidgin (pcm)
|
| 171 |
+
- Ga (gaa)
|
| 172 |
+
- Lozi (loz)
|
| 173 |
+
- Luganda (lug)
|
| 174 |
+
- Gun (guw)
|
| 175 |
+
- Bemba (bem)
|
| 176 |
+
- Efik (efi)
|
| 177 |
+
- Luvale (lue)
|
| 178 |
+
- Luba-Lulua (lua)
|
| 179 |
+
- Tonga (toi)
|
| 180 |
+
- Tshivenḓa (ven)
|
| 181 |
+
- Tumbuka (tum)
|
| 182 |
+
- Tetela (tll)
|
| 183 |
+
- Isoko (iso)
|
| 184 |
+
- Kaonde (kqn)
|
| 185 |
+
- Zande (zne)
|
| 186 |
+
- Umbundu (umb)
|
| 187 |
+
- Mossi (mos)
|
| 188 |
+
- Tiv (tiv)
|
| 189 |
+
- Luba-Katanga (lub)
|
| 190 |
+
- Fula (fuv)
|
| 191 |
+
- San Salvador Kongo (kwy)
|
| 192 |
+
- Baoulé (bci)
|
| 193 |
+
- Ruund (rnd)
|
| 194 |
+
- Luo (luo)
|
| 195 |
+
- Wolaitta (wal)
|
| 196 |
+
- Swazi (ssw)
|
| 197 |
+
- Lunda (lun)
|
| 198 |
+
- Wolof (wol)
|
| 199 |
+
- Nyaneka (nyk)
|
| 200 |
+
- Kwanyama (kua)
|
| 201 |
+
- Kikuyu (kik)
|
| 202 |
+
- Fon (fon)
|
| 203 |
+
- Bambara (bam)
|
| 204 |
+
- Chokwe (cjk)
|
| 205 |
+
- Dinka (dik)
|
| 206 |
+
- Dyula (dyu)
|
| 207 |
+
- Kabyle (kab)
|
| 208 |
+
- Kamba (kam)
|
| 209 |
+
- Kabiyè (kbp)
|
| 210 |
+
- Kanuri (knc)
|
| 211 |
+
- Kimbundu (kmb)
|
| 212 |
+
- Kikongo (kon)
|
| 213 |
+
- Nuer (nus)
|
| 214 |
+
- Sango (sag)
|
| 215 |
+
- Tamasheq (taq)
|
| 216 |
+
- Tamazight (tzm)
|
| 217 |
+
- N'ko (nqo)
|
| 218 |
+
|
| 219 |
+
# Specifically Finetuned on:
|
| 220 |
+
- Amharic (amh)
|
| 221 |
+
- Hausa (hau)
|
| 222 |
+
- Igbo (ibo)
|
| 223 |
+
- Kikuyu (kik)
|
| 224 |
+
- Kinyarwanda (kin)
|
| 225 |
+
- Luo (luo)
|
| 226 |
+
- Twi (twi)
|
| 227 |
+
- Yoruba (yor)
|
| 228 |
+
- Zulu (zul)
|
| 229 |
+
- Ewe (Ewe)
|
| 230 |
+
- Lingala (lin)
|
| 231 |
+
- Wolof (wol)
|