An IR-aided machine learning framework for the BioCreative II.5 Challenge

Area of Science:

• Bioinformatics
• Machine Learning
• Information Retrieval

Background:

• Text mining systems are crucial for extracting biological information.
• Developing efficient and robust text mining frameworks is an ongoing challenge.
• Standardized training data and minimal external resources are desirable for framework development.

Purpose of the Study:

• To develop and evaluate an information retrieval and machine learning framework for text mining.
• To participate in the BioCreative II.5 Challenge tasks using the developed framework.
• To assess the efficiency, robustness, and scalability of the proposed framework.

Main Methods:

• Developed an information retrieval (IR) and machine learning framework.
• Utilized standardized training data with minimal external knowledge and parsing.
• Built text mining systems within the framework for BioCreative II.5 Challenge tasks.

Main Results:

• Systems ranked among the top five teams for raw F1 scores across all three BioCreative II.5 tasks.
• Achieved third place for homonym ortholog F1 scores in the INT task.
• Demonstrated high performance in all evaluated tasks.

Conclusions:

• The developed IR-based framework is efficient and robust for text mining applications.
• The framework shows potential for scalability in handling large biological datasets.
• The approach is effective even with minimal external knowledge and parsing requirements.