boun-tabi/nli_tr

土耳其语自然语言推理（NLI-TR）数据集是通过将SNLI和MNLI这两个基础的NLI语料库使用亚马逊翻译工具翻译成土耳其语而生成的大规模数据集。该数据集包含两个配置：snli_tr和multinli_tr，分别对应SNLI和MNLI的翻译版本。数据集的结构包括训练集、验证集和测试集，每个数据实例包含前提、假设和标签三个字段，标签分为entailment、neutral和contradiction三类。数据集的规模在100K到1M之间，属于单语言（土耳其语）数据集。

原始信息汇总

数据集概述

基本信息

• 语言: 土耳其语（tr）
• 许可证: cc-by-3.0, cc-by-4.0, cc-by-sa-3.0, mit, other
• 多语言性: 单语种
• 大小: 100K<n<1M

数据集组成

• config_name: snli_tr, multinli_tr
• 特征:
  • idx: int32
  • premise: string
  • hypothesis: string
  • label: 分类标签，包括 entailment (0), neutral (1), contradiction (2)
• 数据分割:
  • snli_tr:
    • train: 550152 样本
    • validation: 10000 样本
    • test: 10000 样本
  • multinli_tr:
    • train: 392702 样本
    • validation_matched: 10000 样本
    • validation_mismatched: 10000 样本

数据集创建

• 注释创建者: 专家生成

• 语言创建者: 机器生成

• 源数据集: 扩展自 snli, multi_nli

• 任务类别: 文本分类

• 任务ID: natural-language-inference, semantic-similarity-scoring, text-scoring

• 论文引用:

  @inproceedings{budur-etal-2020-data, title = "Data and Representation for Turkish Natural Language Inference", author = "Budur, Emrah and "{O}zçelik, Rıza and G"{u}ng"{o}r, Tunga", booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)", month = nov, year = "2020", address = "Online", publisher = "Association for Computational Linguistics", abstract = "Large annotated datasets in NLP are overwhelmingly in English. This is an obstacle to progress in other languages. Unfortunately, obtaining new annotated resources for each task in each language would be prohibitively expensive. At the same time, commercial machine translation systems are now robust. Can we leverage these systems to translate English-language datasets automatically? In this paper, we offer a positive response for natural language inference (NLI) in Turkish. We translated two large English NLI datasets into Turkish and had a team of experts validate their translation quality and fidelity to the original labels. Using these datasets, we address core issues of representation for Turkish NLI. We find that in-language embeddings are essential and that morphological parsing can be avoided where the training set is large. Finally, we show that models trained on our machine-translated datasets are successful on human-translated evaluation sets. We share all code, models, and data publicly.", }