Skip to content

Research on neural-symbolic convergence in information retrieval — demonstrating $4.7B industry rediscovering 1985 CDS/ISIS principles

Notifications You must be signed in to change notification settings

capraCoder/neural-symbolic-ir-paper1

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 

Repository files navigation

The Authority of Neural Scale

"The $4.7 billion question has a free answer. It's been waiting in UNESCO archives for forty years."

Research demonstrating that modern billion-dollar vector database systems are unconsciously recreating the architectural principles of UNESCO's 1985 CDS/ISIS system.


Paper

Title: The Authority of Neural Scale
DOI Status: ✅ Ready for submission
Target: Communications of the ACM (Contributed Article)
Document: caprazli-neural-scale.pdf


Abstract

The information retrieval industry is spending $4.7 billion to rediscover a solution distributed for free in 1985. We demonstrate that eight modern systems—FAISS, DiskANN, ScaNN, SPANN, TiDAR, Pinecone, Weaviate, and continuous batching—have independently converged on CDS/ISIS architectural principles.

We propose:

  • Two-Level Theory: Separating semantic understanding (Level 1) from structural organization (Level 2)
  • Neural-to-Symbolic Bridge: Framework projecting 10x cost reduction ($2.4M → $240k/year)

Key Insight

Modern "Innovation" CDS/ISIS Principle (1985)
FAISS Inverted Files Inverted File Indexing
DiskANN Vamana Graph B-tree Navigation
ScaNN Quantization Variable-Length Encoding
SPANN Clustering Field Independence Axiom
Continuous Batching Variable Records (ISO 2709)

Repository Structure

neural-symbolic-ir-paper1/
├── paper1-convergence/
│   ├── manuscript/
│   │   ├── caprazli-neural-scale.tex    # LaTeX source
│   │   ├── caprazli-neural-scale.pdf    # Compiled paper
│   │   ├── references.bib               # 21 references
│   │   └── submission/                  # Submission package
│   └── literature/                      # Research sources
└── shared/
    └── acmart-primary/                  # ACM LaTeX templates

Author

Kafkas M. Caprazli
Independent Researcher, Wolfsburg, Germany
ORCID: 0000-0002-5744-8944

Former member of FAO teams implementing CDS/ISIS-based information systems (AGRIS, CARIS, ASFA, DOCREP). Co-author of research on unified authority files presented at ECDL 2003.


Citation

@article{caprazli2025authority,
  author    = {Caprazli, Kafkas M.},
  title     = {The Authority of Neural Scale},
  journal   = {Communications of the ACM},
  year      = {2025},
  note      = {Submitted December 2025}
}

Acknowledgments

This research represents human-AI collaborative work. The author extensively collaborated with Claude (Anthropic), Perplexity, and Gemini Pro for literature synthesis and drafting. The core insight—that modern systems are unconsciously converging on CDS/ISIS principles—represents the author's original contribution.


License

© 2025 Kafkas M. Caprazli. All rights reserved.

Academic use permitted with citation. For other uses, contact the author.

About

Research on neural-symbolic convergence in information retrieval — demonstrating $4.7B industry rediscovering 1985 CDS/ISIS principles

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages