Scholarly open access journals, Peer-reviewed, and Refereed Journals, Impact factor 8.14 (Calculate by google scholar and Semantic Scholar | AI-Powered Research Tool) , Multidisciplinary, Monthly, Indexing in all major database & Metadata, Citation Generator, Digital Object Identifier(DOI)
The pervasive digitization of institutional records has generated an extensive corpus of scanned, image-encoded PDF documents whose informational content remains inaccessible to conventional text-processing infrastructure. This paper presents and rigorously evaluates an integrated pipeline that unifies multi-engine Optical Character Recognition with semantic retrieval and large language model generation to enable natural language querying over arbitrary PDF documents. The proposed architecture deploys three parallel OCR engines—Tesseract 5.3, EasyOCR 1.7, and PaddleOCR 2.7—with quantitative output selection, followed by LangChain-orchestrated semantic chunking, 384-dimensional dense embedding via the all-MiniLM-L6-v2 sentence transformer, FAISS-indexed vector storage, and Retrieval-Augmented Generation inference over both locally deployed and API-backed large language models. Evaluation across a 42-document, 318-page heterogeneous corpus stratified into three document quality categories yields a FAISS Precision@5 of 0.78 and a Mean Reciprocal Rank of 0.86—representing a 17-percentage-point improvement over TF-IDF baselines. Human-assessed response correctness reached 74% under GPT-3.5-Turbo and 58% under the locally deployed Falcon-RW-1B model, at a mean query latency of 9.5 seconds. PaddleOCR achieved the lowest Word Error Rate of 9.7% on degraded documents, while Tesseract retains a marginal advantage on clean typeset material. The work provides a rigorous framework for OCR engine selection within retrieval-augmented pipelines—a design dimension absent from the prior integrated document intelligence literature.
Keywords:
Optical Character Recognition; Retrieval-Augmented Generation; Document Intelligence; FAISS; Semantic Embeddings; LangChain; PDF Processing; Large Language Models
Cite Article:
"OCR-Integrated Retrieval-Augmented Generation for Intelligent PDF Processing", International Journal for Research Trends and Innovation (www.ijrti.org), ISSN:2456-3315, Vol.11, Issue 5, page no.a796-a806, May-2026, Available :http://www.ijrti.org/papers/IJRTI2605099.pdf
Downloads:
000107
ISSN:
2456-3315 | IMPACT FACTOR: 8.14 Calculated By Google Scholar| ESTD YEAR: 2016
An International Scholarly Open Access Journal, Peer-Reviewed, Refereed Journal Impact Factor 8.14 Calculate by Google Scholar and Semantic Scholar | AI-Powered Research Tool, Multidisciplinary, Monthly, Multilanguage Journal Indexing in All Major Database & Metadata, Citation Generator