Identifying relevant full-text articles for database curation

In this paper, we propose an approach for identifying curatable articles from a large pool. Our system currently considers three parts of an article as three individual representations of the article, and utilizes two domain-specific resources to reveal the deep knowledge contained in the article in order to generate more representations of the article. Cross-validation is employed to find the best combination of representations and an SVM classifier is trained out of this combination. The cross-validation results and results of the official runs are listed. The experimental results show overall high performance.