A Novel Sample Selection Strategy for Imbalanced Data of Biomedical Event Extraction with Joint Scoring Mechanism

Area of Science:

• Bioinformatics
• Computational Biology
• Natural Language Processing

Background:

• Biomedical event extraction is crucial for analyzing vast scientific literature.
• Imbalanced annotated biomedical corpora hinder the performance of classification algorithms.
• Extracting complex, multi-argument events presents significant challenges.

Purpose of the Study:

• To develop an effective method for biomedical event extraction from imbalanced datasets.
• To improve the accuracy of complex event extraction in bioinformatics.
• To address the limitations of existing methods in handling imbalanced biomedical text data.

Main Methods:

• A sample selection algorithm based on sequential patterns was proposed to filter negative training samples.
• Direct extraction of multi-argument events as triplets using a support vector machine classifier.
• A joint scoring mechanism incorporating sentence similarity and trigger importance was employed for result correction.

Main Results:

• The proposed method demonstrated efficient extraction of biomedical events.
• Improved performance in handling imbalanced biomedical text data.
• Successful extraction of complex multi-argument events.

Conclusions:

• The developed sequential pattern-based sample selection and SVM classification approach effectively addresses challenges in biomedical event extraction.
• The joint scoring mechanism further refines the accuracy of extracted events.
• This method offers an efficient solution for extracting complex events from large-scale biomedical literature.