RCB (Russian Commitment Bank)

俄罗斯承诺银行是自然发生的话语的语料库，其最后一句包含在蕴涵取消运算符（问题、模态、否定、条件的先行词）下的子句嵌入谓词。任务类型 RTE (Recognizing Textual Entailment) 句对分类 - 蕴涵 - 矛盾 - 中性 示例 { “前提”：“损失金额为一千卢布。指定调查组前往现场并确定攻击者。他原来是当地居民，以前曾犯过类似的罪行。”，“假设”：“当地居民以前犯过类似的罪行。”，“动词”：“判断”，“否定”：“no_negation” , "label": "entailment", "idx" : 269 } 我们是如何收集数据的？所有文本示例均从开放新闻来源和文学杂志中收集，然后通过 Yandex.Toloka 上的人工评估进行人工审查和补充。

The Russian Commitment Bank Corpus is a collection of naturally occurring utterances, where the final sentence contains a clause-embedding predicate governed by an entailment-canceling operator (including questions, modal expressions, negation, and the antecedent of conditional clauses). The task is Recognizing Textual Entailment (RTE) sentence pair classification, with three categories: entailment, contradiction, and neutral. An example is provided below: { "Premise": "The amount of the loss was one thousand rubles. A designated investigation team was sent to the scene and identified the attacker. He turned out to be a local resident who had previously committed similar crimes.", "Hypothesis": "The local resident had previously committed similar crimes.", "Verb": "judge", "Negation": "no_negation", "label": "entailment", "idx": 269 }. How was the data collected? All text samples were sourced from open news outlets and literary magazines, then manually reviewed and supplemented via human evaluation on Yandex.Toloka.

搜集汇总

背景与挑战

背景概述

RCB（俄罗斯承诺银行）是一个自然语言处理数据集，专注于文本蕴涵识别任务，包含从开放新闻和文学来源收集的文本对，并经过人工标注。该数据集由华为诺亚方舟实验室等机构于2020年发布，用于判断前提和假设之间的蕴涵、矛盾或中性关系。