COVID-19 Literature Compiler : A literature-mining and Data Visualization Tool

Sammanfattning

The number of COVID-19 related articles increased explosively with the start of the pandemic. To save time and effort, researchers use literature-mining tools to find articles efficiently. As COVID-19 related research continues developing, it is important for researchers to know the trend of recent studies. Although there are many advanced literature-mining tools with different designs and using different technologies, most literature-mining tools focus on finding articles by text searching and Natural Language Processing (NLP). Researchers must have specific words in mind and then search for articles by using these words. In this thesis project, a literature-mining tool is built as a web app. This tool uses the relationships among articles, including shared references and Medical Subject Headings (MeSH), to find new related articles and to help researchers further study the literature at hand. This represents a shift from keyword searching to searching based upon relationships among articles and MeSH terms. The main technical problem of the web app is search speed. Different methods, including the use of a graph database and a Single Page App (SPA), were used to improve the app's speed and performance. Another problem is the commonly used MeSH terms for COVID-19 related articles, such as``COVID-19'', ``Humans'', and ``Child'' caused serious noise when finding similar articles that share the same MeSH terms. Different methods such as Over-Representation Analysis (ORA) and Fisher's exact test are used. Different kinds of filters are provided for users to eliminate unwanted result. After the web app was developed, it was tested and the results show that the planned functions could be realized and the search speed is acceptable, i.e. under 3 seconds. However, the issue of MeSH terms leading to broad results remains to be solved.