Learning from positive and unlabeled documents for automated detection of alternative splicing sentences in medline abstracts | IEEE Conference Publication | IEEE Xplore