Gimli

Gimli performs biomedical named-entity recognition to identify and extract biomedical names from scientific literature for biomedical information extraction.

Key Features:

• Comprehensive feature set: Implements orthographic, morphological, linguistic-based, conjunctions, and dictionary-based features for NER.
• Model combination methodology: Provides a method to combine different trained models to leverage multiple model strengths.
• High performance: Reports F-measure values of 87.17% on the GENETAG corpus and 72.23% on the JNLPBA corpus.
• Integration and extensibility: Exposes a library interface and allows extension or adaptation of its functionalities.

Scientific Applications:

• Biomedical named-entity recognition: Identification and extraction of gene, protein, and other biomedical names from scientific text.
• Biomedical information extraction and text mining: Support for bioinformatics and computational biology projects requiring accurate NER as an input to downstream extraction tasks.

Methodology:

Implements multiple feature types (orthographic, morphological, linguistic-based, conjunctions, dictionary-based) and a model combination approach; supports pre-trained models and user-defined training processes.