Dependency-Driven Feature-based Learning for Extracting Protein-Protein Interactions from Biomedical Text

Recent kernel-based PPI extraction systems achieve promising performance because of their capability to capture structural syntactic information, but at the expense of computational complexity. This paper incorporates dependency information as well as other lexical and syntactic knowledge in a feature-based framework. Our motivation is that, considering the large amount of biomedical literature being archived daily, feature-based methods with comparable performance are more suitable for practical applications. Additionally, we explore the difference of lexical characteristics between biomedical and newswire domains. Experimental evaluation on the AIMed corpus shows that our system achieves comparable performance of 54.7 in F1-Score with other state-of-the-art PPI extraction systems, yet the best performance among all the feature-based ones.

• Type: Journal
• Year: 2010
• Where: COLING
• Authors: Bing Liu, Longhua Qian, Hongling Wang, Guodong Zhou
• Keywords: COLING 2010, Comparable Performance, Computational Linguistics, Kernel-based Ppi Extraction, PPI Extraction Systems