Working Paper
Mixed Methods: Performance of OCR, Fuzzy Search, and LLMs with Image-Based Text
Abstract: Over time, humans have created vast archives of unstructured textual data that remain largely inaccessible to contemporary researchers. Advances in optical character recognition (OCR), full-text search, and large language models (LLMs) offer unprecedented opportunities to unlock insights from historical source material by making it technically and economically feasible to analyze these data. We present a framework to consider the various costs and trade-offs of different methods for working with a large corpus of images containing unstructured text. We illustrate this framework and present performance statistics via a case study identifying racial restrictions in 5.1 million images of Philadelphia property deeds from 1910–1965. We find 8,395 deed images containing a racial restriction, which we estimate to cover more than 6% of all parcels in Philadelphia.
JEL Classification: C81; G5; N32; R31; R38;
https://doi.org/10.21799/frbp.wp.2026.54
Access Documents
File(s): File format is application/pdf https://www.philadelphiafed.org/-/media/frbp/assets/working-papers/2026/wp26-54.pdf
Bibliographic Information
Provider: Federal Reserve Bank of Philadelphia
Part of Series: Working Papers
Publication Date: 2026-11-16
Number: 26-54