abstract
- Processing huge repository of medical literature for extracting relevant and high-quality evidences demands efficient evidence support methods. We aim at developing methods to automate the process of finding quality evidences from a plethora of literature documents and grade them according to the context (local condition). We propose a two-level methodology for quality recognition and grading of evidences. First, quality is recognized using quality recognition model; second, context-aware grading of evidences is accomplished. Using 10-fold cross-validation, the proposed quality recognition model achieved an accuracy of 92.14 percent and improved the baseline system accuracy by about 24 percent. The proposed context-aware grading method graded 808 out of 1354 test evidences as highly beneficial for treatment purpose. This infers that around 60 percent evidences shall be given more importance as compared to the other 40 percent evidences. The inclusion of context in recommendation of evidence makes the process of evidence-based decision-making "situation-aware."