Grafos de conocimiento científico. Análisis comparativo de publicaciones mediante un enfoque multidimensional

Abstract

Abstract: Scientific Knowledge Graphs describe research resources or objects, such as scientific documents, use of semantic technologies, ontologies and metadata. The goal of SKG is to build large amounts of data to improve the search for information. This project proposes a multi-dimensional model for the analysis of documents, indexing and similarity metrics. The model will allow researchers to have different perspectives or views to find similar documents, by considering different metadata. To validate the SKG, an enriched data was prepared with: (1) publication metadata obtained from Scholarly Data and (2) entities and triplets which are stated in the abstract. The prototype used Python notebooks and libraries to work graphs like SPARQLWrapper, RDFLib and Network, and Natural Language Processing as Spacy. Additionally, we identify DBpedia entities and SKOS concepts with the TagMe API. Finally, we analyze documents by using the similarity of GraphDB creating semantic indexes with SPARQL to visualize Network and the Pyvis library were used. The results show thatout of twenty manually analyzed publications we found twenty-four relevant publications, that is, the system was able to find at least one similar document of interest to the researcher.