File size: 1,385 Bytes
24e5b4e aeebad0 24e5b4e aeebad0 24e5b4e aeebad0 24e5b4e 5083d70 4c41635 5083d70 24e5b4e 650b1b7 24e5b4e ca4aa0e 24e5b4e 6cee675 24e5b4e 0ab3892 24e5b4e f2601a7 24e5b4e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 |
---
language:
- en
- hu
tags:
- translation
license: apache-2.0
metrics:
- sacrebleu
- chrf
widget:
- text: >-
This may not make much sense to you, sir, but I'd like to ask your
permission to date your daughter.
---
# mT5 Translation model
For further models, scripts and details, see [our repository](https://github.com/nytud/machine-translation) or [our demo site](https://juniper.nytud.hu/demo/nlp).
- Source language: English
- Target language: Hungarian
- Pretrained model used: mT5-small
- Finetuned on subcorpora from OPUS
- Segments: 56.837.602
## Limitations
- tokenized input text (tokenizer: [HuSpaCy](https://huggingface.co/huspacy))
- max_source_length = 128
- max_target_length = 128
## Results
| Model | BLEU | chrF-3 | chrF-6 |
| ------------- | ------------- | ------------- | ------------- |
| Google en-hu | 25.30 | 54.08 | 49.06 |
| BART | 36.89 | 60.77 | 56.4 |
| **mT5** | **27.69** | **53.73** | **48.57** |
## Citation
If you use this model, please cite the following paper:
```
@inproceedings {laki-yang-mt,
title = {{Jobban fordítunk magyarra, mint a Google!}},
booktitle = {XVIII. Magyar Számítógépes Nyelvészeti Konferencia},
year = {2022},
publisher = {Szegedi Tudományegyetem, Informatikai Intézet},
address = {Szeged, Magyarország},
author = {Laki, László and Yang, Zijian Győző},
pages = {357--372}
}
``` |