Related Experiment Video
Updated: Apr 30, 2026
05:32
High-speed Video Microscopy Analysis for First-line Diagnosis of Primary Ciliary Dyskinesia
Published on: January 19, 2022
4.6K
A novel feature selection strategy for enhanced biomedical event extraction using the Turku system
Jingbo Xia1, Alex Chengyu Fang2, Xing Zhang2
1College of Science, Huazhong Agricultural University, Wuhan, Hubei 430070, China ; Department of Chinese, Translation and Linguistics, City University of Hong Kong, Kowloon, Hong Kong.
Biomed Research International
|May 7, 2014
Summary
This study optimized the Turku Event Extraction System (TEES) by refining its high-dimensional feature set using an accumulated effect evaluation (AEE) algorithm. This improved performance in text-mining tasks, enhancing the system
Area of Science:
- Natural Language Processing
- Bioinformatics
- Machine Learning
Background:
- High-dimensional features are crucial for text-mining classifiers.
- The Turku Event Extraction System (TEES) excels in bio-medical text mining but relies on numerous features.
- Optimizing feature selection is key to enhancing TEES performance.
Purpose of the Study:
- To analyze the contribution of individual feature classes within the TEES.
- To identify and modify the feature set for improved text-mining accuracy.
- To enhance the overall performance of the Turku Event Extraction System.
Main Methods:
- Implementation of an accumulated effect evaluation (AEE) algorithm.
- Application of a greedy search strategy for feature analysis.
- Systematic evaluation of feature class contributions in TEES.
Main Results:
- Identified key features contributing to TEES performance.
- Modified the feature set based on AEE analysis.
- Achieved an increased F-score from 51.21% to 53.27% (strict criteria) for Task 1.
- Reached 57.24% performance using approximate span and recursive criteria.
Conclusions:
- Feature selection significantly impacts text-mining classifier performance.
- The refined feature set enhances the Turku Event Extraction System.
- Optimized TEES demonstrates improved accuracy in bio-medical event extraction.
