SimConcept

SimConcept identifies and normalizes composite named entities in biomedical text, resolving individual concepts within composite spans such as "BRCA1/2" to support named entity recognition (NER) and normalization.

Key Features:

• Hybrid approach: Integrates machine learning with pattern identification strategies and rule-based pattern recognition to identify components within composite mentions.
• Composite entity resolution: Identifies and resolves multitype composite named entities that span multiple concepts.
• Coverage beyond coordination ellipsis: Handles composite forms beyond simple coordination ellipsis commonly found in biomedical literature.
• Performance: Reports F-measures of 90.42% for gene identification and resolution, 86.47% for diseases, and 86.05% for chemicals.
• Improved concept recognition: Improves overall performance of concept recognition and normalization for genes and diseases.

Scientific Applications:

• Named entity recognition and normalization: Supports NER and normalization tasks on biomedical text containing composite mentions.
• Gene identification and resolution: Facilitates gene identification and resolution in cases of composite mentions (e.g., "BRCA1/2").
• Disease and chemical concept recognition: Supports disease and chemical concept recognition with reported F-measures.
• Biomedical literature extraction: Enables more precise data extraction and analysis in genomics, pathology, and chemical biology.

Methodology:

Combines machine learning models with pattern identification strategies and rule-based pattern recognition to identify and resolve individual components within composite named entity spans.