Skip to main content

Dataset & Benchmark Results

Paper: DocEng Short Paper - Cost-Aware Human-LLM Collaboration for Post-OCR Corrections
Authors: Stergios Konstantinidis, Hayman Lotfy, Michalis Vlachos (University of Lausanne)
Venue: ACM Symposium on Document Engineering (DocEng) - Short Paper
GitHub: https://github.com/Stergios-Konstantinidis/Cost-Aware-Human-LLM-Collaboration-for-post-OCR-Corrections
Overleaf: https://git.overleaf.com/6a2cb93de1bec86186ee8613

Experimental Benchmark & Evaluation

1. Archival Corpus

Evaluated across 609 text segments spanning 45 digitized pages from 9 Swiss French-language historical periodicals (Canton of Vaud, 1733–1945):

Periodical Era Issues Pages Segments Characteristics
La Revue 1875–1945 4 5 139 19th/20th century serif, multi-column
Feuille d'Avis 1762–1841 4 6 118 Antique typography, long 's' ligatures
Mercure Suisse 1733–1748 3 5 92 18th-century early modern French
Gazette de Lausanne 1804–1920 5 8 145 High density, complex layout
Others (Tribune, etc.) 1880–1930 5 21 115 Mixed commercial & political press

2. Key Benchmark Results

Error Rate Comparison (Tesseract OCR Base):

  • Raw OCR Baseline: $CER = 6.30%$30\%, $WER = 14.20%$20\%
  • All-LLM Baseline (GPT-4o / Gemini 1.5 Flash): $CER = 2.92%$92\% (Cost: 100% token budget)
  • Selective Pre-Routing (Ours): $CER = 3.12%$12\% (Bypasses 40% of all segments!)
  • Collaborative Safeguard (<5% Human Budget): $CER = 2.49%$49\% (14% relative reduction over All-LLM)

Comparison with Routing Baselines:

  1. Random Routing: Weakest; frequently sends clean text and degrades it.
  2. OCR Confidence (Thresholding): Underperforms because Tesseract confidence scores poorly correlate with semantic error in archaic fonts.
  3. ConfBERT: Good detection but computationally heavy.
  4. Our Regression Router + Safeguard: Consistently establishes the Pareto frontier across all budget limits.

Experimental Evaluation & Visual Benchmark Plots

Historical Newspaper Segmentation Sample

[![

doceng_segment_example.png

Figure: doceng_segment_example.png](https://wiki.stergios.ch/uploads/images/gallery/2026-09/scaled-1680-/doceng-segment-example.png)](https://wiki.stergios.ch/uploads/images/gallery/2026-09/doceng-segment-example.png)png

Figure 1: Representative historical archival newspaper segment illustrating layout artifacts and OCR degradation.

CER Reduction Across Correction Strategies

[![

correction_strategies_comparison.png

Figure: correction_strategies_comparison.png](https://wiki.stergios.ch/uploads/images/gallery/2026-09/scaled-1680-/correction-strategies-comparison.png)](https://wiki.stergios.ch/uploads/images/gallery/2026-09/correction-strategies-comparison.png)png

Figure 2: Character Error Rate (CER) reduction across baseline OCR, All-LLM, ConfBERT, and Collaborative Routing.

OCR Engine Improvement Comparison

[![

engine_improvement_chart.png

Figure: engine_improvement_chart.png](https://wiki.stergios.ch/uploads/images/gallery/2026-09/scaled-1680-/engine-improvement-chart.png)](https://wiki.stergios.ch/uploads/images/gallery/2026-09/engine-improvement-chart.png)png

Figure 3: Relative error reduction across historical OCR engines (Tesseract, Abbyy, and modern HTR).

Strategy Performance Heatmap

[![

engine_strategy_heatmap.png

Figure: engine_strategy_heatmap.png](https://wiki.stergios.ch/uploads/images/gallery/2026-09/scaled-1680-/engine-strategy-heatmap.png)](https://wiki.stergios.ch/uploads/images/gallery/2026-09/engine-strategy-heatmap.png)png

Figure 4: Comprehensive performance heatmap comparing LLM prompt strategies across historical periodicals.

Back to top