Identifying Protein-Protein Interaction Sentences Using Boosting and Kernel Methods

As the amount of biological research literature increases, flnding information is becoming a daunting task. Since machine learning techniques could alleviate this problem, we propose a ma- chine learning framework to identify protein-protein interaction sentences from research papers. This machine learning technique is one of the basic components needed to automatically extract biological information from texts. Since the protein-protein interaction (PPI) sentences have their own patterns at article and sentence levels, these patterns are mined by using boosting and kernel methods. Both approaches have good characteristics for the PPI extraction tasks, and naturally can handle heuristic information for future extensions.