Duda Stephany, Aliferis Constantin, Miller Randolph, Statnikov Alexander, Johnson Kevin
Department of Biomedical Informatics, Vanderbilt University, Nashville, TN, USA.
AMIA Annu Symp Proc. 2005;2005:216-20.
Drug-drug interaction systems exhibit low signal-to-noise ratios because of the amount of clinically insignificant or inaccurate information they contain. MEDLINE represents a respected source of peer-reviewed biomedical citations that potentially might serve as a valuable source of drug-drug interaction information, if relevant articles could be pinpointed effectively and efficiently. We evaluated the classification capability of Support Vector Machines as a method for locating articles about drug interactions. We used a corpus of "positive" and"negative" drug interaction citations to generate datasets composed of MeSH terms, CUI-tagged title and abstract text, and stemmed text words. The study showed that automated classification techniques have the potential to perform at least as well as PubMed in identifying drug-drug interaction articles.
药物相互作用系统由于包含大量临床意义不大或不准确的信息，其信噪比很低。医学文献数据库（MEDLINE）是经过同行评审的生物医学文献的权威来源，如果能够有效且高效地找到相关文章，它有可能成为药物相互作用信息的宝贵来源。我们评估了支持向量机作为一种定位药物相互作用相关文章的方法的分类能力。我们使用了一个由“阳性”和“阴性”药物相互作用文献引用组成的语料库，来生成由医学主题词（MeSH）、带概念唯一标识符（CUI）标签的标题和摘要文本以及词干文本单词组成的数据集。研究表明，自动分类技术在识别药物相互作用文章方面至少有潜力与PubMed表现得一样好。
