An Automated Recommendation System for Back-of-the-Book Index (BoBI)

Abstract

The Indexing is the structured key terms and concepts within a document; it allows the readers to find particular information in the document quickly, efficiently, and correctly without needing to search the text as a whole. Existing indexing approaches face several limitations. It was done manually with the help of skilled indexers through well known applications, particularly when dealing with large information. Current systems used to select text one by one manually through software, or extract keywords by software, which often misses indexes keys already done by others and domain-specific terms. One promising direction involves the utilization of prior index datasets or libraries, where historical index entries are reused to improve efficiency and consistency in new documents. An interactive interface is required to allow indexers to review, accept, reject, or modify the recommended terms. To address these situations, this research proposes an Automated Index Recommendation System (AIRS) that streamlines the design book index through a structured three-phase approach, Input, Process, and Output. By bridging automation with indexer validation, the proposed system not only reduces workload, but also checks accuracy, consistency, and scalability. The proposed framework contributes to forward publishing and resource management by providing an efficient and reliable solution for modern needs. The results show a F1-score greater than 0.97, the filtering function redundant, the search accuracy 98%, and the HIR remained 98%, and the output achieved a precision of 99.5%.