Text Categorization of Biomedical Data Sets Using Graph Kernels and a Controlled Vocabulary

摘要: Recently, graph representations of text have been showing improved performance over conventional bag-of-words in categorization applications. In this paper, we present a graph-based representation for biomedical articles and use kernels to classify those into high-level categories. our representation, common concepts semantic relationships are identified with the help an existing ontology used build rich structure that provides consistent feature set preserves additional information could improve classifier's performance. We attempt graphs using both set-based kernel is capable dealing disconnected nature simple linear kernel. Finally, report results comparing classification classifiers text-based classifiers.