A Systematic Framework For Structured Document Classification And Decision Support Systems Leveraging Question Paper–Syllabus, Text Book Alignment Datasets

Objectives: To demonstrate the rapid expansion of digital academic content such as textbooks, syllabi, and question papers has created a need for automated systems that can reliably classify exam questions according to subject and syllabus unit and determine the associated page number from the textbook. Method: This study proposes a novel Syllabus Question Mapping (SQM) filtering model integrated with a Bag-of-Words (BoW) representation and Naïve Bayes classification to identify the appropriate subject and unit for a given question. The framework was implemented using Jupyter Notebook with Python libraries. The dataset consists of textbooks and question papers collected from online academic sources. Questions were extracted from multiple file formats (PDF, DOCX, etc.) and manually verified to create the gold standard. In the SQM filtering stage, keywords are extracted from syllabus headings, unit descriptions, and textbook content to construct a curated unit-wise keyword dictionary. The SQM filter performs direct keyword matching to identify candidate units by measuring keyword overlap between unit content and the given question. The filtered features are then used to train a Naïve Bayes classifier, which performs probabilistic classification to predict the most appropriate subject and unit label. Questions with no valid match are classified as “Out of Syllabus.” Findings: Experimental evaluation across multiple subjects shows that the Proposed (Syllabus Question Mapping) SQM-Naïve Bayes framework consistently outperforms traditional classifiers such as Logistic Regression, SVM, and KNN, reaching classification accuracy between 98% and 100%. Novelty/Applications: The major contribution lies in the two-stage hybrid filtering and probabilistic classification approach, enabling accurate syllabus-unit mapping with minimal training data. Beyond academia, the SQM framework can be extended to structured document classification and section mapping in domains such as banking, healthcare, libraries, recruitment, and human resource management. The results confirm that the proposed system is an efficient and reliable solution for automated academic question classification and syllabus mapping. Keywords: Syllabus, Question, Classification, Naive Bayes, BoW