Contextual label sensitive gated network for biomedical event trigger extraction.

Area of Science:

• Biomedical informatics
• Natural Language Processing
• Computational Biology

Background:

• Biomedical event extraction is vital for advancing biomedical research.
• Accurate event trigger identification is a critical prerequisite for event extraction pipelines.
• Previous methods overlook contextual label associations and require dynamic feature weight adjustment.

Purpose of the Study:

• To propose a novel contextual label sensitive gated network for biomedical event trigger extraction.
• To address limitations in existing methods by dynamically mixing features and capturing contextual label clues.
• To enhance the representation of semantic information using dependency-based word embeddings and attention mechanisms.

Main Methods:

• Developed a contextual label sensitive gated network.
• Integrated dependency-based word embeddings to capture semantic information.
• Employed an attention mechanism for focused representations.
• Dynamically adjusted the weighting of word embeddings and contextual features.

Main Results:

• The proposed network effectively captures contextual label clues and dynamically mixes features.
• Achieved state-of-the-art performance on the Multi-Level Event Extraction (MLEE) corpus.
• Demonstrated superior F1-score compared to existing methods.

Conclusions:

• The novel network significantly advances biomedical event trigger extraction.
• The approach provides a more robust and accurate method for identifying biomedical events.
• This work contributes to improved biomedical research through enhanced information extraction.