TeamTat

TeamTat facilitates collaborative manual annotation of biomedical text to produce BioC-format corpora for development and evaluation of text-mining and information-extraction algorithms.

Key Features:

• Multi-User Collaboration: Supports team annotation with project manager roles, independent annotator workspaces, annotator assignment, and anonymous distribution of documents to mitigate bias.
• Input Formats and Retrieval: Accepts plain text, PDF, and BioC input files and can ingest documents uploaded locally or automatically retrieved from PubMed/PMC.
• Annotation Schema Management: Allows project managers to define annotation schemas for specific entities and relations within documents.
• Figure Integration: Displays figures extracted from full-text documents alongside text for annotation.
• Quality Assurance Metrics: Computes inter-annotator agreement statistics for corpus quality assessment.
• BioC Output: Exports annotated documents in BioC format with inline annotations for downstream text-mining applications.

Scientific Applications:

• Text-mining and information-extraction algorithm development: Produces manually annotated BioC corpora for training and evaluating NLP algorithms.
• Biomedical NLP corpus creation and benchmarking: Supports entity and relation annotation schemas and inter-annotator agreement metrics to generate gold-standard datasets.
• Literature curation and figure–text linking: Ingests PubMed/PMC content and PDFs and includes figure display to support comprehensive curation of full-text articles.

Methodology:

Documents are ingested from local uploads or PubMed/PMC in plain text, PDF, or BioC formats; project managers specify annotation schemas, assign annotators, and distribute documents (optionally anonymously); annotators work in independent workspaces; inter-annotator agreement statistics are computed and annotated output is exported in BioC with inline annotations.