Natural language processing, especially in the field of social media analysis

该项目将开发一个开放集错误信息分类模型，解决错误信息中的几个挑战。所提出的模型将在开集（即无限数量的错误信息类别）假设下设计，这提供了鲁棒的类别分类性能与不断变化的错误信息类别。该模型还将提供新的类别检测功能和预测解释，以帮助事实核查人员和相关专业人员发现错误信息主题的变化。该项目还将研究少量学习方法，以允许快速模型适应错误信息类别的演变。

This project will develop an open-set misinformation classification model, which addresses several challenges in misinformation. The proposed model will be designed under an open-set (i.e. infinite number of misinformation categories) assumption, which provides robust category classification performance with continuously evolving misinformation categories. The model will also provide new category detection functionality with prediction explanation to help fact-checkers and relevant professionals discover shifts in misinformation topics. The project will also investigate few-shot learning approaches to allow fast model adaptation in response to misinformation category evolution.