The dataset is made up of WikiHow articles. We propose a unified multi-class active learning approach for automatically labeling articles. The experimental results show that the proposed approach works effectively even with a significantly reduced amount of labeled data.
The aim is to build a multi-class classifier of Wikihow pages. A Wikihow page pertains to one and only one of 19 macro-categories. Right now resources are labeled by hand by the creator of the resource, but our aim is to build an automatic tool for the task, based on the article's text and summary.
The report can be found here.
The dataset is made up of WikiHow articles. We propose a unified multi-class active learning approach for automatically labeling articles. The experimental results show that the proposed approach works effectively even with a significantly reduced amount of labeled data.
https://github.com/MarshaGomez/Multilabel-Classification