Systematic review using a semi-supervised bibliometric methodology for application in a precipitation process

The growth in the amount of papers in scientific databases have a detrimental impact on searches by increasing the browsing time and complexity of the queries. There is also a growing need for upskilling current search practices so they provide answers beyond query matches. A specialized and structured document retrieval system was developed to accomplish a systematic review in chemical engineering. Bibliographically coupled citation networks, commonly studied in Scientometrics, provided the framework for modelling the knowledge structure of vaterite research. The retrieval system was integrated by a classification function for document selection followed by a ranking algorithm. The ranking function used network centrality measures for further tuning of the retrieved documents. The classification function consisted of a document tracing method implemented on top of the community detection procedure to approach the optimization of modularity as a statistical classification problem. The goal of the pearl set was to define document relevance during clustering. Accuracy was the main evaluation metric under study. A community with 170 papers on the synthesis of vaterite using the spontaneous precipitation method was found and a total of 256 experiments and 36 variables were collected. The novel approach provides some evidence that search systems can be engineered in ways not considered before.