NewsPH-NLI

NewsPH-NLI是首个针对低资源语言菲律宾语的句子蕴含基准数据集，由马尼拉德拉萨大学创建。该数据集通过自动从菲律宾主要新闻网站收集的229,571篇新闻文章中生成，最终包含600,000个前提-假设对。创建过程利用了新闻文章的结构特点，如单句段落和倒金字塔结构，自动生成蕴含关系。该数据集主要用于评估和改进在低资源语言环境下的自然语言处理模型，特别是Transformer模型的性能。

NewsPH-NLI is the first natural language inference (NLI) benchmark dataset for the low-resource language Filipino, created by De La Salle University Manila. It is generated from 229,571 news articles automatically collected from major Philippine news websites, ultimately containing 600,000 premise-hypothesis pairs. The dataset construction process leverages the structural characteristics of news articles such as single-sentence paragraphs and the inverted pyramid structure to automatically generate entailment relations. This dataset is primarily used to evaluate and improve the performance of natural language processing models in low-resource language scenarios, with a particular focus on Transformer models.

搜集汇总

背景与挑战

背景概述

NewsPH-NLI是首个针对低资源语言菲律宾语的句子蕴含基准数据集，由马尼拉德拉萨大学创建，包含60万个前提-假设对，通过自动从新闻文章中生成。该数据集利用新闻文章的结构特点，旨在评估和改进低资源语言环境下的自然语言处理模型性能，特别是Transformer模型。