DrNote

DrNote annotates and links biomedical entities in unstructured medical and clinical texts using OpenTapioca, WikiData, and Wikipedia to support information extraction across languages.

Key Features:

• Entity linking (OpenTapioca): Entity linking is performed using OpenTapioca to map mentions to knowledge base entries.
• Knowledge bases (WikiData, Wikipedia): Uses WikiData and Wikipedia datasets for entity resolution and contextual information.
• Multilingual support: Supports English and non-English languages and can train on language-specific Wikipedia datasets.
• Raw text processing: Processes unstructured raw text inputs such as clinical notes and medical reports.
• PDF processing (Tesseract OCR): Integrates Tesseract OCR to extract text from PDF documents for downstream annotation.
• Customizable annotation scope: Allows selection of specific subsets of the knowledge base to tailor annotated entity types.
• Electronic health reports handling: Targets extraction of relevant information from electronic health reports and similar clinical documents.

Scientific Applications:

• Clinical trials text mining: Extraction and linking of entities from clinical trial documents to support trial analysis and curation.
• Medical research: Annotation of biomedical entities in research corpora to enable information retrieval and knowledge discovery.
• Electronic health record analysis: Entity extraction from EHRs and clinical notes to surface clinically relevant information.
• Multilingual biomedical studies: Application to studies requiring entity linking and annotation across multiple languages using language-specific Wikipedia data.

Methodology:

Entity linking is performed with OpenTapioca using WikiData and Wikipedia datasets (including language-specific Wikipedia datasets for training), raw text and PDF inputs are processed with optional Tesseract OCR, and the annotation scope is configured by selecting subsets of the knowledge base.