带反馈的多标签文本分类

multi-label text classification with feedback

我目前正在研究多标签多class 文本class化问题。 在问题的某些部分,我需要向用户提供反馈,即算法根据哪些特征、单词或短语决定 class将实例归入某个类别?

例如,对于像 A B C D E F G H I J K L M N O P Q R 这样的文本,我需要这样的输出:

我该如何考虑这个问题?这个问题的名称或关键字是什么?是否有针对此类问题的现成 python 软件包?

您正在寻找的是机器学习 (ml) 模型中的可解释性。有很多 discussion on this topic and is steadily gaining more attention, as more real-world applications in important subjects such as terrorism detection, precision-medicine 等合并 ml 组件。

至少有一个 python 模块已经在做你似乎要求的事情,名为 lime

如果您总体上对任务感兴趣,您还可以查看许多其他资源(例如 extended presentation and article, conferences 等)。