Predicting the factuality of news reporting and bias of media outlets is surely relevant for automated news credibility and fact-checking. While prior work has focused on the veracity of news, we propose a fine-grained reliability analysis of the entire media. Specifically, we study the prediction of sentence-level factuality of news reporting and bias of media outlets, which may explain more accurately the overall reliability of the entire source. We first manually produced a large sentence-level dataset, titled "FactNews", composed of 6,191 sentences expertly annotated according to factuality and media bias definitions from AllSides. As a result, baseline models for sentence-level factuality prediction were presented by fine-tuning BERT. Finally, due to the severity of fake news and political polarization in Brazil, both dataset and baseline were proposed for Portuguese. However, our approach may be applied to any other language.

翻译：预测新闻报道的事实性和媒体机构的倾向性，无疑对自动新闻可信度评估与事实核查具有重要意义。此前研究主要关注新闻的真实性，而本文则提出对整体媒体进行细粒度的可靠性分析。具体而言，我们研究了新闻句子级别的事实性预测以及媒体机构的倾向性，这能够更准确地解释整个信息来源的整体可靠性。我们首先手动构建了一个大规模句子级数据集“FactNews”，包含6,191个句子，这些句子根据AllSides机构对事实性和媒体倾向的定义进行了专家标注。在此基础上，通过微调BERT模型，提出了句子级事实性预测的基线模型。最后，鉴于巴西虚假新闻和政治极化的严重性，我们为葡萄牙语提供了相应的数据集和基线模型。然而，本方法可适用于其他任何语言。