Publishing machine actionable reproducible scholarly knowledge

Abstract

Scientific research faces many challenges related to the credibility of published results. In essence, there is typically not enough documentation on how experiments are conducted and data is generated. Thus, increasing the reliability of articles through reproducibility will improve the quality of the published scientific literature and others better reliable results. This thesis describes today's problem of the research literature related to non-reproducibility and unstructured data such as weak experiments designs, errors, data dredging and under-specified methods. We suggest a variety of solutions to resolve these problems through linking machine readability with the reproducibility of the information in academic papers. We use therefore a knowledge platform which provides reproducibility on one side and on the other side another platform that ensures the machine actionability of data. Then, we build an integration between them and test it on a selected use case article. After establishing the integration, we obtained, as a result, a reproducible article described in machine-actionable and structured manner. Thereafter, we created a solution that allow every reader to switch between the static and dynamic (reproducible and machine-readable) form of the article. This thesis discusses the benefits and limitations of these observed results and emphasizes the future alternatives.