XML to HTML Conversion Showcase

This page provides examples of TEI XML documents that have been transformed into static HTML pages using Python and Jupyter Notebooks.

The text is based on OCR output from the Hathi Trust. I have corrected errors as I have seen them in an opportunistic fashion. These files are all works in progress that are, I hope, useful for many purposes now but all can benefit from more work. That said, the quality of the base OCR-output from Hathi (which receives its OCR from Google) has gotten remarkably good. Until c. 2023, mixed Greek and Roman script was awful -- OCR found it oddly difficult to switch between scripts. Now, the base OCR for Greek + Roman script is (assuming good scans and clear print) an excellent start.

I used Gemini Pro 2.5 to convert the raw OCR-generated text into TEI XML. I found that it was best to give a very general prompt asking for TEI XML because Gemini often had better ideas about how to structure the output that I would have suggested, because, after decades of work, my instinct is to produce something relatively simple. I learned quite a bit working with Gemini. Gemini generated the tables and paradigms in Pharr, for example, from the raw OCR text. I am astonished at how well it did. Some tables need correction. Only one table was too complex to yield useful results (table 1 on pp. lx-lxi)

Likewise, Gemini did a very solid job capturing the tabular format of the Russell Latin-Greek grammar.

In each case, I have used LLMs (so far mainly Gemini 2.5) to write Python programs to go from TEI XML to HTML. The Parallel Latin-Greek Grammar and Pharr's Homeric Greek have some relatively complicated interactivity.