数据集:
bbc_hindi_nli
数据集为印地语。
'train' 的一个例子如下所示。
{'hypothesis': 'यह खबर की सूचना है|', 'label': 'entailed', 'premise': 'गोपनीयता की नीति', 'topic': '1'}
转换过程的源数据集为 BBC 印地语头条新闻数据集( https://github.com/NirantK/hindi2vec/releases/tag/bbc-hindi-v0.1 )。
请参考这篇论文:" https://www.aclweb.org/anthology/2020.aacl-main.71" "。
注释过程在 "数据集创建" 部分已经描述。
谁是注释者?注释是自动完成的。
数据集中没有提到个人和敏感信息。
请参考这篇论文:" https://www.aclweb.org/anthology/2020.aacl-main.71 "。
请参考这篇论文:" https://www.aclweb.org/anthology/2020.aacl-main.71 "。
没有其他已知限制。
请参考这个链接:" https://github.com/midas-research/hindi-nli-data "。
仓库中写道:" https://github.com/avinsit123/hindi-nli-data "
版权所有 (C) 2019 年德里多模数字媒体分析实验室 - 新德里印度信息技术学院 (MIDAS, IIIT-Delhi)。请联系作者获取有关数据集的任何信息。
@inproceedings{uppal-etal-2020-two, title = "Two-Step Classification using Recasted Data for Low Resource Settings", author = "Uppal, Shagun and Gupta, Vivek and Swaminathan, Avinash and Zhang, Haimin and Mahata, Debanjan and Gosangi, Rakesh and Shah, Rajiv Ratn and Stent, Amanda", booktitle = "Proceedings of the 1st Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 10th International Joint Conference on Natural Language Processing", month = dec, year = "2020", address = "Suzhou, China", publisher = "Association for Computational Linguistics", url = "https://www.aclweb.org/anthology/2020.aacl-main.71", pages = "706--719", abstract = "An NLP model{'}s ability to reason should be independent of language. Previous works utilize Natural Language Inference (NLI) to understand the reasoning ability of models, mostly focusing on high resource languages like English. To address scarcity of data in low-resource languages such as Hindi, we use data recasting to create NLI datasets for four existing text classification datasets. Through experiments, we show that our recasted dataset is devoid of statistical irregularities and spurious patterns. We further study the consistency in predictions of the textual entailment models and propose a consistency regulariser to remove pairwise-inconsistencies in predictions. We propose a novel two-step classification method which uses textual-entailment predictions for classification task. We further improve the performance by using a joint-objective for classification and textual entailment. We therefore highlight the benefits of data recasting and improvements on classification performance using our approach with supporting experimental results.", }
感谢 @avinsit123 添加了这个数据集。