Concept Graph Preserving Semantic Relationship for Biomedical Text Categorization

Nowadays, graph illustration of text is gaining importance owing to improved performance over traditional bag-of-words representations in text categorization applications. During this paper, we tend to have a graph-based illustration for biomedical articles and use graph kernels to classify those articles into high level classes. During this approach, common biomedical concepts and linguistics relationships are identified with the help of an existing ontology and are used to build a chic graph structure that has a regular feature set and preserves extra linguistics info that would improve a classifier's performance. We tend to classify the graphs victimisation each a set-based graph kernel that's capable of coping with the disconnected nature of the graphs and an easy linear kernel. Finally, we tend to report the results scrutiny the classification performance of the kernel classifiers to common text based classifiers.