nilc-nlp/assin
收藏资源简介:
ASSIN(Avaliação de Similaridade Semântica e INferência textual)语料库是一个包含葡萄牙语句子对的语料库,适用于文本推理和释义分类器的探索。该语料库包含从葡萄牙和巴西的新闻文章中提取的句子对,分别使用欧洲葡萄牙语(EP)和巴西葡萄牙语(BP)编写。数据集的创建过程包括从Google News中收集描述同一事件的新闻文章,并使用LDA模型检索相似句子对。最终的数据集包含10,000个句子对,其中一半为巴西葡萄牙语,另一半为欧洲葡萄牙语。每个语言变体包含2,500个训练对、500个验证对和2,000个测试对。
ASSIN (Avaliação de Similaridade Semântica e INferência textual) corpus is a Portuguese sentence pair corpus intended for the exploration of textual inference and paraphrase classifiers. This corpus includes sentence pairs extracted from news articles of Portugal and Brazil, written in European Portuguese (EP) and Brazilian Portuguese (BP) respectively. The dataset was constructed by collecting news articles covering the same event from Google News, and retrieving similar sentence pairs using LDA models. The final dataset contains 10,000 sentence pairs, with half in Brazilian Portuguese and the other half in European Portuguese. For each language variant, there are 2,500 training pairs, 500 validation pairs, and 2,000 test pairs.
数据集概述
名称: ASSIN (Avaliação de Similaridade Semântica e INferência textual)
语言: 葡萄牙语 (pt)
许可证: 未知
多语言性: 单语种
大小: 10K<n<100K
源数据集: 原始数据
任务类别: 文本分类
任务ID:
- 文本评分
- 自然语言推理
- 语义相似性评分
配置:
- full: 默认配置,包含5000个训练样本,1000个验证样本,4000个测试样本。
- ptbr: 包含2500个训练样本,500个验证样本,2000个测试样本。
- ptpt: 包含2500个训练样本,500个验证样本,2000个测试样本。
数据集结构
特征:
sentence_pair_id: int64premise: stringhypothesis: stringrelatedness_score: float32entailment_judgment: class_label (0: NONE, 1: ENTAILMENT, 2: PARAPHRASE)
数据分割:
- full:
- 训练: 5000样本, 986499字节
- 测试: 4000样本, 767304字节
- 验证: 1000样本, 196821字节
- ptbr:
- 训练: 2500样本, 463505字节
- 测试: 2000样本, 374424字节
- 验证: 500样本, 91203字节
- ptpt:
- 训练: 2500样本, 522994字节
- 测试: 2000样本, 392880字节
- 验证: 500样本, 105618字节




