Transductive Support Vector Classification for RNA Related Biological Abstracts

Support Vector Machines use a set of related supervised learning methods for classification and regression. When used for classification, the SVM algorithm creates a hyper plane that separates the data into two classes with the maximum- margin. Given positive and negative training examples a maximum-margin hyper plane is identified where it splits the positive from the negative examples, while maximizing the margin. Transductive Inference enhances the learning process by attempting to achieve the lowest error rate possible given a small sample of training examples. In this research we developed a set of software tools to convert scientific abstracts into support vectors that could be used with an implementation of Support Vector Machine called SVM-Light to classify the abstracts. Three distinct classification experiments were conducted: First, to classify abstracts about RNA research out of a set of randomly selected Abstracts. Second, to classify abstracts about specific types of RNA research out of a set of abstracts that all contain the expression "RNA." Third, to classify tRNA, mRNA, snRNA, and rRNA abstracts individually out of a set of abstracts pertaining to the four categories of RNA.