Constructing and evaluating automated literature review systems

Network Analysis and Natural Language Processing to Obtain a Landscape of the Scientific Literature on Materials Applications

Recent progress in natural language processing (NLP) enables mining the literature in various tasks akin to knowledge discovery. Obtaining an updated birds-eye view of key research topics and their evolution in a vast, dynamic field such as materials science is challenging even for experienced researchers. In this Perspective paper, we present a landscape of the area of applied materials in selected representative journals based on a combination of methods from network science and simple NLP strategies. We found a predominance of energy-related materials, e.g., for batteries and catalysis, organic electronics, which include flexible sensors and flexible electronics, and nanomedicine with various topics of materials used in diagnosis and therapy. As for the impact calculated through standard metrics of impact factor, energy-related materials and organic electronics are again top of the list across different journals, while work in nanomedicine has been found to have a lower impact in the journals analyzed. The adequacy of the approach to identify key research topics in materials applications was verified indirectly by comparing the topics identified in journals with diverse scopes, including journals that are not specific to materials. The approach can be employed to obtain a fast overview of a given field from the papers published in related scientific journals, which can be adapted or extended to any research area.

Personalized and Self-Management: Systematic Search and Evaluation Quality Factors and User Preference of Drug Reference Apps in Taiwan

Background: Drug reference apps promote self-management and improve the efficiency and quality of work for physicians, nurses, pharmacists, and patients. This study aimed to describe a systematic and stepwise process to identify drug reference apps in Taiwan, assess the quality of these apps, and analyze the influential factors for user ratings. Methods: A two-step algorithm (KESS) consisting of keyword growing and systematic search was proposed. Seven independent reviewers were trained to evaluate these apps using Mobile App Rating Scale (MARS). A logistic regression model was fitted and average marginal effects (AME) were calculated to identify the effects of factors for higher user ratings. Results: A total of 23 drug reference apps in Taiwan were identified and analyzed. Generally, these drug reference apps were evaluated as acceptable quality with an average MARS score of 3.23. Higher user engagement, more functionality, better aesthetics, and more information associated with higher user ratings. Navigation is the most influential factor on higher user ratings (AME: 13.15%) followed by performance (AME: 11.03%), visual appeal (AME: 10.87%), credibility (AME: 10.67%), and quantity of information (AME: 10.42%). Conclusions: User experience and information clearly affect user ratings of drug reference apps. Five key factors should be considered when designing drug reference apps.