five

frtna/es_it_Results-base-OPUS_Tatoeba

收藏
Hugging Face2022-01-04 更新2024-03-04 收录
下载链接:
https://hf-mirror.com/datasets/frtna/es_it_Results-base-OPUS_Tatoeba
下载链接
链接失效反馈
官方服务:
资源简介:
- Model: [OPUS-MT](https://huggingface.co/Helsinki-NLP/opus-mt-es-it) - Tested on: [Tatoeba]() <br> - Metric: - bleu(tensorflow), - sacrebleu(github->mjpost), - google_bleu(nltk), - rouge(google-research), - meteor(nltk), - ter(university of Maryland) <br> - Retrieved from: [Huggingface](https://huggingface.co/metrics/) [metrics](https://github.com/huggingface/datasets/blob/master/metrics/) - Script used for translation and testing: [https://gitlab.com/hmtkvs/machine_translation/-/tree/production-stable](https://gitlab.com/hmtkvs/machine_translation/-/tree/production-stable) ## Info ## mtdata-OPUS Tatoeba (length=14178, single reference) **bleu** : 0.5228 <br> **sacrebleu** : 0.5652 <br> **google_bleu** : 0.5454 <br> **rouge-mid** : precision=0.7792, recall=0.7899, f_measure=0.7796 <br> **meteor** : 0.7557 <br> **ter** : score=0.3003, num_edits= 24654, ref_length= 82079.0 ## OPUS Tatoeba (length = 5000, multi references) **bleu** : 0.5165 <br> **sacrebleu** : 0.7098 <br> **google_bleu** : 0.5397 <br> **rouge-mid** : precision=0.9965, recall=0.5021, f_measure=0.6665 <br> **meteor** : 0.3344 <br> **ter** : score: 0.6703, 'num_edits': 38883, 'ref_length': 58000.0

- 模型:OPUS-MT,链接:https://huggingface.co/Helsinki-NLP/opus-mt-es-it - 测试数据集:Tatoeba - 评估指标: - 基于TensorFlow实现的双语评估替补(Bilingual Evaluation Understudy,简称BLEU) - sacreBLEU(由GitHub用户mjpost实现) - 基于NLTK工具包的谷歌官方BLEU - 由谷歌研究院开发的ROUGE(Recall-Oriented Understudy for Gisting Evaluation)指标 - 基于NLTK工具包的METEOR(Metric for Evaluation of Translation with Explicit ORdering)指标 - 由马里兰大学开发的翻译编辑率(Translation Edit Rate,简称TER) - 数据来源:Hugging Face指标库(https://huggingface.co/metrics/)与Hugging Face数据集官方指标仓库(https://github.com/huggingface/datasets/blob/master/metrics/) - 翻译与测试所用脚本:https://gitlab.com/hmtkvs/machine_translation/-/tree/production-stable ## 信息 ### mtdata-OPUS Tatoeba(样本量=14178,单参考译文) **BLEU**:0.5228 **sacreBLEU**:0.5652 **谷歌官方BLEU**:0.5454 **ROUGE-MID**:精确率=0.7792,召回率=0.7899,F1值=0.7796 **METEOR**:0.7557 **TER**:得分=0.3003,总编辑操作数=24654,参考译文总长度=82079.0 ### OPUS Tatoeba(样本量=5000,多参考译文) **BLEU**:0.5165 **sacreBLEU**:0.7098 **谷歌官方BLEU**:0.5397 **ROUGE-MID**:精确率=0.9965,召回率=0.5021,F1值=0.6665 **METEOR**:0.3344 **TER**:得分=0.6703,总编辑操作数=38883,参考译文总长度=58000.0
提供机构:
frtna
原始信息汇总

数据集概述

数据集名称

  • mtdata-OPUS Tatoeba (length=14178, single reference)
  • OPUS Tatoeba (length = 5000, multi references)

评估指标

  • bleu
  • sacrebleu
  • google_bleu
  • rouge
  • meteor
  • ter

评估结果

mtdata-OPUS Tatoeba (length=14178, single reference)
  • bleu : 0.5228
  • sacrebleu : 0.5652
  • google_bleu : 0.5454
  • rouge-mid : precision=0.7792, recall=0.7899, f_measure=0.7796
  • meteor : 0.7557
  • ter : score=0.3003, num_edits= 24654, ref_length= 82079.0
OPUS Tatoeba (length = 5000, multi references)
  • bleu : 0.5165
  • sacrebleu : 0.7098
  • google_bleu : 0.5397
  • rouge-mid : precision=0.9965, recall=0.5021, f_measure=0.6665
  • meteor : 0.3344
  • ter : score: 0.6703, num_edits: 38883, ref_length: 58000.0
5,000+
优质数据集
54 个
任务类型
进入经典数据集
二维码
社区交流群

面向社区/商业的数据集话题

二维码
科研交流群

面向高校/科研机构的开源数据集话题

数据驱动未来

携手共赢发展

商业合作