COVID-19 Kaggle Literature Organization

Maksim Ekin Eren, Nick Solovyev, Edward Raff, Charles Nicholas, Ben Johnson

Abstract

The world has faced the devastating outbreak of Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2), or COVID-19, in 2020. Research in the subject matter was fast-tracked to such a point that scientists were struggling to keep up with new findings. With this increase in the scientific literature, there arose a need for organizing those documents. We describe an approach to organize and visualize the scientific literature on or related to COVID-19 using machine learning techniques so that papers on similar topics are grouped together. By doing so, the navigation of topics and related papers is simplified. We implemented this approach using the widely recognized CORD-19 dataset to present a publicly available proof of concept.

Introduction

In the early months of 2020, the scientific community faced an unprecedented "infodemic." Research on SARS-CoV-2 was being published at a staggering rate—over 23,000 papers by May 2020, with the volume doubling every 20 days. This massive influx of information left health professionals and researchers overwhelmed. Traditional search methods were insufficient for discovering latent connections between papers or getting a "bird's-eye view" of the research landscape. The authors sought to build a tool that could automatically group similar papers together and visualize these relationships, making it easier to navigate the literature without manual tagging.

Methods

The study utilized the CORD-19 (COVID-19 Open Research Dataset), which contained over 141,000 scholarly articles. After cleaning (removing duplicates, non-English papers, and entries without full text), the authors focused on a corpus of 49,967 papers.

The authors built a multi-stage NLP and machine learning pipeline:

  1. Pre-processing: Cleaning and tokenization using SciSpacy, a specialized biomedical parser.
  2. Vectorization: Conversion of text into numerical feature vectors using TF-IDF (Term Frequency-Inverse Document Frequency).
  3. Dimensionality Reduction: Principal Component Analysis (PCA) was applied to reduce the feature space while retaining 95% of the variance.
  4. Clustering: k-means clustering was performed on the PCA-reduced vectors. To find the optimal number of clusters ($k$), the "elbow method" was used.
Elbow method plot showing distortion vs. k-value, identifying k=20 as optimal.

Key Results

The "elbow curve" analysis determined that 20 clusters provided a balanced organization of the CORD-19 dataset. For the final 2D map, the authors used t-SNE (t-distributed Stochastic Neighbor Embedding), a non-linear dimensionality reduction technique designed to preserve local structures in high-dimensional data.

t-SNE visualization of the 20 clusters identified in the CORD-19 dataset.

Biomedical engineers from Utah State University validated the tool by successfully scanning for specific drug effectiveness and broad-spectrum antivirals, confirming that the tool grouped related, high-value research.

Discussion

The analysis revealed that terminology significantly drives clustering. "Niche" topics, such as animal infection, formed tight clusters due to specific, consistent terminology. Conversely, broader topics like epidemiology were more spread out, sharing vocabulary across multiple clusters. Interestingly, even though t-SNE and k-means were applied independently, they largely agreed on cluster boundaries, suggesting a robust thematic structure in the literature. Furthermore, the pipeline demonstrated cross-domain utility, as it was successfully adapted to cluster literature on tropical reef fish and plastic pollution.

Limitations and Future Work

Projecting high-dimensional data into 2D inevitably loses some structural information; some overlapping clusters might be distinct in higher dimensions. The choice of $k=20$ is a heuristic; higher values might reduce distortion but split meaningful clusters. Future work should focus on dynamic tools capable of handling the continuous, daily influx of new research.

Summary updated

Abstract

The world has faced the devastating outbreak of Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2), or COVID-19, in 2020. Research in the subject matter was fast-tracked to such a point that scientists were struggling to keep up with new findings. With this increase in the scientific literature, there arose a need for organizing those documents. We describe an approach to organize and visualize the scientific literature on or related to COVID-19 using machine learning techniques so that papers on similar topics are grouped together. By doing so, the navigation of topics and related papers is simplified. We implemented this approach using the widely recognized CORD-19 dataset to present a publicly available proof of concept.
BibTeX
@inproceedings{eren2020covid,
  author = {Maksim Ekin Eren and Nick Solovyev and Edward Raff and Charles Nicholas and Ben Johnson},
  title = {{COVID-19 Kaggle Literature Organization}},
  booktitle = {DocEng},
  year = {2020},
  url = {https://openreview.net/forum?id=w6CU0hd0bSo},
}
Figure