Rule-Based Classification and Clustering for Thematic Analysis of Research Papers

The rapid expansion of scientific publications has intensified the need for faster and more consistent approaches to organizing research evidence. This work presents a domainindependent, keyword-driven framework for the automatic categorization and thematic grouping of research articles using only titles and abstracts. The framework combines rule-based labeling, expert validation, and unsupervised clustering to support transparent and reproducible analysis of publication trends. To demonstrate its use, the approach is applied to the Three-Way Clustering literature as a case study. The workflow assigns preliminary technology-oriented or application-oriented labels using a simple keyword model. Expert review refines these labels to form the final ground truth. The validated abstracts are then grouped into coherent themes using TF-IDF and Kmeans. The case study illustrates how the process can summarize methodological directions and identify the balance between technology-oriented and application-oriented research. Because the framework does not rely on training data or domain-specific models, it can be adapted to other research fields by redefining the keyword sets and applying the same automated workflow.