TeamTat: a collaborative text annotation tool

Area of Science:

• Biomedical Informatics
• Computational Biology
• Natural Language Processing

Background:

• Manual data annotation is crucial for developing text-mining and information-extraction algorithms in the rapidly expanding biomedical literature.
• Existing annotation tools often lack features for figure display, project management, and collaborative team annotation, hindering efficiency and quality.

Purpose of the Study:

• To develop and present TeamTat, a novel web-based annotation tool designed to facilitate efficient and high-quality multi-user, multi-label document annotation for biomedical research.
• To address the limitations of current tools by providing comprehensive support for the entire annotation lifecycle, from project setup to quality assessment.

Main Methods:

• TeamTat is a web-based platform with a local setup option, supporting multi-user, multi-label document annotation.
• Features include customizable annotation schemas, anonymous document distribution, various input formats (text, PDF, BioC), and BioC output.
• The tool integrates figure display, independent user workspaces, project management dashboards, and inter-annotator agreement statistics for quality control.

Main Results:

• TeamTat effectively manages the entire annotation production lifecycle for teams.
• It supports diverse document input and BioC output formats, displays figures, and enables anonymous distribution to prevent bias.
• The platform facilitates collaboration, task tracking, and corpus quality assessment through inter-annotator agreement metrics and review interfaces.

Conclusions:

• TeamTat offers a robust solution for managing complex, multi-user annotation projects in the biomedical domain.
• Its features enhance efficiency, ensure expert quality, and support collaborative efforts in creating high-value annotated datasets for text-mining applications.