Inferring global-scale temporal latent topics from news reports to predict public health interventions for COVID-19
Bibliographic record
Abstract
Abstract The COVID-19 global pandemic has highlighted the importance of non-pharmacological interventions (NPI) for controlling epidemics of emerging infectious diseases. Despite the importance of NPI, their implementation has been monitored in an ad hoc and uncoordinated manner, mainly through the manual efforts of volunteers. Given the absence of systematic NPI tracking, authorities and researchers are limited in their ability to quantify the effectiveness of NPI and guide decisions regarding their use during the progression of a global pandemic. To address this issue, we propose 3-stage machine learning framework called EpiTopics to facilitate the surveillance of NPI by mining the vast amount of unlabelled news reports about these interventions. Building on topic modeling, our method characterizes online government reports and media articles related to COVID-19 as a mixture of latent topics. Our key contribution is the use of transfer-learning to address the limited number of NPI-labelled documents and topic modelling to support interpretation of the results. At stage 1, we trained a modified version of the unsupervised dynamic embedded topic model (DETM) on 1.2 million international news reports related to COVID-19. At stage 2, we used the trained DETM to infer topic mixture from a small set of 2000 NPI-labelled WHO documents as the input features for predicting NPI labels on each document. At stage 3, we supply the inferred country-level temporal topics from the DETM to the pretrained document-level NPI classifier to predict country-level NPIs. We identified 25 interpretable topics, over 4 distinct and coherent COVID-related themes. These topics contributed to significant improvements in predicting the NPIs labelled in the WHO documents and in predicting country-level NPIs. Together, our work lay the machine learning methodological foundation for future research in global-scale surveillance of public health interventions. The EpiTopics code is available at GitHub: https://github.com/li-lab-mcgill/covid-npi.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.012 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".