MedTAG: a portable and customizable annotation tool for biomedical documents.

Area of Science:

• Biomedical Informatics
• Natural Language Processing

Background:

• Developing Named Entity Recognition and Linking (NER+L) algorithms in the biomedical domain is hindered by a lack of sufficient annotated data.
• Manual annotation by experts is expensive and time-consuming, limiting the progress of NLP in this field.

Purpose of the Study:

• To introduce MedTAG, a novel collaborative biomedical annotation tool.
• To support, organize, and accelerate the manual annotation process for biomedical documents.

Main Methods:

• MedTAG is presented as an open-source, platform-independent, and freely distributable tool.
• Its features are demonstrated through its application in annotating over seven thousand clinical reports in histopathology.
• A comparative analysis of MedTAG against established tools like BioQRator, ezTag, MyMiner, and tagtog is performed.

Main Results:

• MedTAG was successfully used by physicians and experts to manually annotate a substantial volume of clinical reports.
• The tool demonstrated a user-friendly interface and efficient annotation capabilities.
• MedTAG was highlighted as one of the few open-source tools with an open license and easy cross-platform installation.

Conclusions:

• MedTAG was designed adhering to key requirements for manual annotation tools, including availability, distributability, installability, workability, and schematic design.
• The tool meets a significant majority (20 out of 22) of criteria from a recent review of annotation tools.
• MedTAG offers a valuable, efficient, and accessible solution for biomedical data annotation.