Cross-Lingual Contradiction Detection
收藏资源简介:
本研究提出了一种面向结构的跨语言矛盾检测数据集,旨在评估不同深度学习模型在自然语言推理任务中的表现。数据集通过应用结构依赖规则自动生成句子对,涵盖了多种语言结构和逻辑形式,如布尔协调、量词、确定描述和计数操作。该数据集支持英语和葡萄牙语,用于比较模型的跨语言性能,并诊断模型在处理结构推理时的能力。数据集的应用领域包括提升聊天机器人等实际应用中的逻辑一致性。
This study proposes a structure-oriented cross-lingual contradiction detection dataset, aiming to evaluate the performance of diverse deep learning models in natural language inference tasks. The dataset automatically generates sentence pairs by applying structure-dependent rules, covering multiple linguistic structures and logical forms such as Boolean coordination, quantifiers, definite descriptions, and counting operations. Supporting English and Portuguese, this dataset is designed to compare the cross-lingual performance of models and diagnose their capabilities in handling structural reasoning. Its application scenarios include improving logical consistency in practical applications such as chatbots.

- 1A logical-based corpus for cross-lingual evaluation计算机科学系,数学与统计研究所,圣保罗大学,巴西 · 2019年



