Coreference resolution improves extraction of Biological Expression Language statements from texts

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Bioinformatics

Background:

• Automated extraction of biological events from scientific literature is crucial for knowledge discovery.
• Biological Expression Language (BEL) provides a standardized format for representing biological knowledge.
• Integrating coreference resolution with event extraction can potentially enhance the accuracy of biological information retrieval.

Purpose of the Study:

• To develop and evaluate a system for automatically extracting biological events and translating them into BEL statements.
• To investigate the impact of coreference resolution on biological event extraction in the biomedical domain.
• To assess the system's performance in the BioCreative V BEL track.

Main Methods:

• The system utilizes existing text mining components for coreference resolution and biological event extraction.
• A novel strategy for BEL statement generation was incorporated and formally tested.
• The system was evaluated on the BioCreative V (2015) BEL track (Track 4).

Main Results:

• The system achieved the best F-score performance of 20.2 (stage 1) and 35.2 (stage 2) for full BEL statement level.
• Evaluation on the training dataset demonstrated benefits from integrating coreference resolution with event extraction.
• The system successfully translated extracted biological events into BEL statements.

Conclusions:

• The developed system effectively automates the extraction of biological events and their translation into BEL statements.
• Coreference resolution integration positively impacts biological event extraction accuracy.
• The system shows strong performance in BEL statement generation for biomedical literature analysis.