TrigNER: automatically optimized biomedical event trigger recognition on scientific documents.

Area of Science:

• Biomedical text mining
• Computational biology
• Bioinformatics

Background:

• Cellular events are crucial for understanding biological processes and disease mechanisms.
• Automated extraction of these events from literature accelerates biomedical knowledge updates.
• Identifying event trigger words is a critical, yet challenging, step in event extraction pipelines.

Purpose of the Study:

• To develop an automated solution for recognizing biomedical event triggers.
• To address challenges in feature selection, context representation, and event type classification for trigger words.

Main Methods:

• Utilized Conditional Random Fields (CRFs) with an extensive feature set, including linguistic, orthographic, morphological, local context, and dependency parsing features.
• Implemented a configurable algorithm for automatic optimization of feature sets and training parameters for each event type.
• Developed optimized CRF models tailored to the linguistic characteristics of specific event types.

Main Results:

• TrigNER achieved a 62.7 F-measure on the BioNLP 2009 shared task corpus.
• Outperformed existing methods in recognizing specific event triggers such as gene expression, transcription, protein catabolism, phosphorylation, and binding.
• Demonstrated effective automatic selection of informative features and optimization of CRF model parameters.

Conclusions:

• TrigNER offers researchers an accessible method for applying advanced techniques in biomedical event trigger recognition.
• The tool simplifies and expedites event recognition from scientific articles, facilitating hypothesis generation and knowledge discovery.
• The open-source solution is available for the biomedical text mining community.