A Historical Analysis of the Field of OR/MS using Topic Models

This study investigates the content of the published scientific literature in the fields of operations research and management science (OR/MS) since the early 1950s. Our study is based on 80,757 published journal abstracts from 37 of the leading OR/MS journals. We have developed a topic model, using Latent Dirichlet Allocation (LDA), and extend this analysis to reveal the temporal dynamics of the field, journals, and topics. Our analysis shows the generality or specificity of each of the journals, and we identify groups of journals with similar content, which are both consistent and inconsistent with intuition. We also show how journals have become more or less unique in their scope. A more detailed analysis of each journals’ topics over time shows significant temporal dynamics, especially for journals with niche content. This study presents an observational, yet objective, view of the published literature from OR/MS that would be of interest to authors, editors, journals, and publishers. Furthermore, this work can be used by new entrants to the fields of OR/MS to understand the content landscape, as a starting point for discussions and inquiry of the field at large, or as a model for other fields to perform similar analyses.

💡 Research Summary

This paper presents a comprehensive historical analysis of the intellectual content within the fields of Operations Research and Management Science (OR/MS) spanning six decades from 1952 to 2012. The core methodology employs Latent Dirichlet Allocation (LDA), a probabilistic topic modeling technique, to analyze the textual content of 80,757 article abstracts collected from 37 leading journals in the field.

The research process began with data acquisition and preprocessing, where non-content articles (e.g., editorials, announcements) were filtered out. The LDA model was then trained on the corpus, configured with 40 topics (K=40). This model assumes each document (abstract) is a mixture of these latent topics, and each topic is a probability distribution over words. While more complex models like Dynamic Topic Models were considered, standard LDA was chosen for computational feasibility given the dataset size. The primary output—the topic distribution for each individual document—was subsequently aggregated in three distinct ways to enable multi-faceted analysis: by year (to see field-wide trends), by journal (to understand journal identity), and by both journal and year (to trace the evolution of specific journals).

The findings reveal several key insights. First, at the field level, the topic composition has diversified over time, moving from a period dominated by a few core topics to a more pluralistic landscape with many specialized research streams. Second, analysis at the journal level clearly distinguishes between “generalist” journals (e.g., Operations Research, Management Science, European Journal of Operational Research), which cover a broad spectrum of topics, and “niche” journals (e.g., Queueing Systems, Journal of Global Optimization), which exhibit high concentration on one or two specific topics. Third, clustering journals based on their topic profiles revealed both intuitive and non-intuitive groupings, suggesting shared methodological foundations beyond traditional subject classifications. For instance, Transportation Research Part B and Journal of Scheduling showed similarity, likely due to a common focus on mathematical optimization.

A particularly significant finding concerns temporal dynamics. While generalist journals showed relatively stable topic distributions, niche journals displayed pronounced volatility, with rapid rises and falls in the prominence of their core topics. This suggests that specialized sub-fields are more susceptible to research trends, paradigm shifts, or technological disruptions. Furthermore, the analysis indicates that many journals have become either more unique or more generalized in their scope over the decades, reflecting strategic editorial decisions or natural evolution of the field’s sub-disciplines.

The authors position this work as an observational, data-driven map of the OR/MS literature. It serves as a valuable resource for authors selecting publication venues, editors and publishers defining journal scope, and new entrants seeking to understand the historical and contemporary landscape of the field. Beyond its immediate application to OR/MS, the study provides a replicable methodological framework for conducting large-scale, content-based historical analyses in other scientific disciplines.