Supporting HIV literature screening with data sampling and supervised learning

摘要: This paper presents a supervised learning approach to support the screening of HIV literature. The manual biomedical literature is an important task in process systematic reviews. Researchers and curators have very demanding, time-consuming error-prone manually identifying documents that must be included review concerning specific problem. We implemented tasks, by automatically flagging potentially selected list retrieved database search. To overcome main issues associated with automatic task, we evaluated use data sampling, feature combinations, selection methods, generating total 105 classification models. models yielding best results were composed Logistic Model Trees classifier, fairly balanced training set, combination Bag-Of-Words MeSH terms. According our results, system correctly labels great majority relevant documents, it could used reviews allow researchers assess greater number less time.