Hilberg's law as learning a world state: implications and limits for text compression

Hilberg's hypothesis states that block entropy of text grows as hn + An^beta. We connect maximal entropy random walk (Parry) sequence ensembles, ANS-like equipartition of surprisal, Santa Fe processes with an HCR-motivated Gaussian world model, and LLM-based entropy measurements into a picture of text as progressive revelation of a world state. Calibration on Santa Fe processes shows that PPMd compression slopes barely respond to the true Hilberg exponent, while a word-level Pitman-Yor coder recovers it but measures lexicon occupancy. A permutation-calibrated lexical index detects word-order effects in anthologies. For compression, a self-contained coder cannot remove a genuine An^beta source term; explicit word and recency models reduce predictive log loss by 1.5-3.4% over a context-mixing base, and shared dictionaries pay off mainly for small independently coded documents. Supplementary files contain all code and results.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-12
DOI
https://doi.org/10.5281/zenodo.24153421
Primary Topic
Algorithms and Data Compression
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Hilberg's law as learning a world state: implications and limits for text compression

Jarek Duda
Zenodo (CERN European Organization for Nuclear Research)
Algorithms and Data Compression
article

Hilberg's law as learning a world state: implications and limits for text compression

Jarek Duda
article en

Abstract

Hilberg's hypothesis states that block entropy of text grows as hn + An^beta. We connect maximal entropy random walk (Parry) sequence ensembles, ANS-like equipartition of surprisal, Santa Fe processes with an HCR-motivated Gaussian world model, and LLM-based entropy measurements into a picture of text as progressive revelation of a world state. Calibration on Santa Fe processes shows that PPMd compression slopes barely respond to the true Hilberg exponent, while a word-level Pitman-Yor coder recovers it but measures lexicon occupancy. A permutation-calibrated lexical index detects word-order effects in anthologies. For compression, a self-contained coder cannot remove a genuine An^beta source term; explicit word and recency models reduce predictive log loss by 1.5-3.4% over a context-mixing base, and shared dictionaries pay off mainly for small independently coded documents. Supplementary files contain all code and results.

Zenodo (CERN European Organization for Nuclear Research)
Algorithms and Data Compression
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon