Research map / Vision-language and multimodal / Human and embodied modeling
American Stories: A Large-Scale Structured Text Dataset of Historical U.S. Newspapers
Melissa Dell, Jacob Carlson, Tom Bryan, Emily Silcock, Abhishek Arora, Zejiang Shen, Luca D'Amico-Wong, Quan Le, Pablo Querubin, Leander Heldring
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| apply lightweight spellchecker after OCR | American Stories pipeline · ten labeled scans | CER: .051 -> .044 | Improved |
| apply spellchecker to headlines | American Stories pipeline · headlines | CER: .089 -> .092 | Hurt |
| use neural classifier instead of keyword methods | topic classification · article level | F1: 58.1 -> 83.6 | Improved |
| use structured American Stories instead of Chronicling America page OCR for neural classification | RoBERTa classifier · page level topic | F1: 83.3 -> 96.0 | Improved |
| use American Stories for sparse mining at page level | keyword mining · page level topic | F1: 83.7 -> 82.8 | No clear effect |
| use neural method instead of Viral Texts at page level | reproduced content detection · page level | ARI: 74.6 -> 86.2 | Improved |
| apply Viral Texts to American Stories instead of Chronicling America | Viral Texts · page level | ARI: 71.4 -> 74.6 | Improved |
| use EffOCR-Word (Small) instead of EffOCR-C (Base) | EffOCR · Chronicling America textlines | CER: 0.023 -> 0.015 | Improved |
| use EffOCR-Word instead of Tesseract | OCR engine · Chronicling America textlines | CER: 0.106 -> 0.015 | Improved |
| use EffOCR-Word instead of Google Cloud Vision | OCR engine · Chronicling America textlines | CER: 0.005 -> 0.015 | Hurt |
| fine-tune TrOCR (Base) on Chronicling America | TrOCR (Base) · day-per-decade | CER: 0.038 -> 0.027 | Improved |
| train TrOCR (Base) from scratch on target data | TrOCR (Base) · Chronicling America textlines | CER: 0.015 -> 0.809 | Hurt |
Closest papers by meaning
- Newswire: A Large-Scale Structured Database of a Century of Historical News (NeurIPS 2024 · similarity 0.85)
- OmniCorpus: A Unified Multimodal Corpus of 10 Billion-Level Images Interleaved with Text (ICLR 2025 · similarity 0.81)
- Position: On the Possibilities of AI-Generated Text Detection (ICML 2024 · similarity 0.80)
- Reference-Based Post-OCR Processing with LLM for Precise Diacritic Text in Historical Document Recognition (AAAI 2025 · similarity 0.80)
- Towards Unified Multi-granularity Text Detection with Interactive Attention (ICML 2024 · similarity 0.80)
- DocLayLLM: An Efficient Multi-modal Extension of Large Language Models for Text-rich Document Understanding (CVPR 2025 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
