A kernel-based SVM for semantic relations extraction from biomedical literature.

Recognising and extracting semantic relationships among named entities; relation extraction is a significant methodology for knowledge representation. In order to capture the semantic as well as syntactic structures in text and to enable deep understanding of biomedical literature, relation extraction becomes essential. The automatic extraction of disease-gene relations is presented in this paper by utilising shallow linguistic features of global and local word sequence context with string kernel-based support vector machine (SVM) for efficient disease-gene relation extraction. The performance of the proposed work shows that the bag-of-features kernel-based SVM classification is a promising resolution for specific disease-gene association mining. The initial results obtained using shallow linguistic kernel methods on an annotated Huntington disease corpora suggested the global tri-grams context surrounding related entities are critical and essential for disease-gene relation extraction, which is in the pact with PPI relation extraction evaluation using AImed corpora.