An Article-Oriented Framework for Automatic Semantic Analysis of COVID-19 Researches

ABSTRACT

In this work, we propose an article-oriented framework for automatically extract semantic topics from scientific articles related to COVID-19 researches. Our framework has four key building blocks (i) pre-processing, (ii) topic modeling, (iii) correlating topics, authors, and institutions, and (iv) a summarization interface. The first one corresponds to apply traditional textual pre-processing strategies in the texts extracted from articles and constructing their data representations. The topic modeling block aims at finding semantic topics from articles based on the most relevant words of each discovered topic. The third block correlates these discovered topics with the articles, authors (researchers), and institutions. The summarization interface provides an intuitive visualization for these results. Our evaluation shows that our framework can automatically extract relevant features from the articles, identifying the key topics covered by them, as well as the contribution of researchers, institutions, and countries to the topics. Our framework can help research institutions and companies to form multidisciplinary teams and funding agencies to identify more promising research approaches regarding COVID-19. © 2021, Springer Nature Switzerland AG.