From reviews to risk assessment
A text mining approach to subclassify injury narratives in online reviews
Bibliographic Data
| ID | 16757315 |
|---|---|
| Authors | Nohel Zaman (0000-0002-4893-0042, San Diego State University), David M Goldberg (0000-0003-2617-5143, San Diego State University, corresponding author), Alan S Abrahams (0000-0002-4884-3192, Virginia Tech) |
| Year | 2026 |
| Volume | 29 |
| Issue | 4 |
| Pages | 1-16 |
| Publication date | 2026-04-17 |
| Peer Reviewed | Yes |
| Open Access | No |
| Type | ARTICLE |
| Venue | Journal of Risk Research (JOURNAL) |
| Journal identifiers | ISSN: 1366-9877 • E-ISSN: 1466-4461 |
| Publisher | Informa UK Limited (PUBLISHER • GB) |
| DOI | 10.1080/13669877.2026.2656748 |
| OpenAlex | W7154727276 |
| Language | EN |
| References cited | 35 |
Businesses and regulators benefit from rapidly detecting, categorizing, and prioritizing injury narratives across product categories. Online reviews are a valuable data source for identifying such narratives, but the challenge lies in extracting the most useful information from the enormous volume of data. In this study, we discover and categorize injury reports in online reviews using key terms prevalent in subcategories of injury narratives. We employ various text classification methods to delineate injury concerns across diverse product and service categories. Additionally, we test the generalizability of our methodologies on online reviews collected from two different data sources: Amazon product reviews and a Fortune 50 retailer’s online reviews. Our semi-automated classification procedure successfully subcategorizes safety concerns by injury type. We show a statistically significant improvement in performance over other competing sentiment-based methodologies. The methodologies used demonstrated effective performance in identifying and categorizing injury reports across both datasets. This approach enhances the efficiency of monitoring and addressing safety concerns. Managerially, the information helps prioritize areas in greatest need of product safety improvement. Additionally, the capacity to identify injury attributes enables policymakers and injury prevention experts to rapidly monitor specific concerns and adjust policies or resources to better serve the public
Narrative · Peer review · Risk communication · Social media · Text messaging · Text mining · Biomedical Text Mining and Ontologies · Mental Health via Writing · Sentiment Analysis and Opinion Mining
The use of the area under the ROC curve in the evaluation of machine learning algorithms
The Measurement of Observer Agreement for Categorical Data
Descriptive framework of injury data
Risk identification in Japanese consumer product injury data throughout an ontology-based knowledge base
Bringing safety analytics to the online shopper
| Citation velocity | historical |
|---|---|
| Highly cited | No |