Dealing with highly imbalanced textual data gathered into similar classes | IEEE Conference Publication | IEEE Xplore