Tagtog is a web-based framework for the annotation of named entities. The tagtog system allows curators to annotate terms manually and leverages machine learning methods to perform the same type of annotations computationally. Initially, the tool is trained with a small set of manually annotated documents. tagtog can then be used to process a set of novel documents wherein automatically generated predictions are made, which can be reviewed and validated by the user. This continuous and interactive retraining of the machine learning methods with user feedback can lead to an ever-improving performance in automatic prediction. Once optimized, the trained machine learning methods can be used to process and annotate a large volume of documents to a sufficiently accurate level.
This description has first appeared in this journal article.