Find misspelled words

For a single large HTML ebook the most reliable route is: strip out markup and run automated checks against a controlled dictionary, then review the flagged items manually. That keeps code, attributes and example snippets from polluting results and avoids dangerous global edits that change your HTML structure (a point @MidiMagic rightly warned about).

Suggested workflow:

1. Extract visible text only (skip script, style, pre, code and attribute values) using an HTML-aware extractor such as Beautiful Soup. See the Beautiful Soup docs.
2. Run a spell/grammar checker that accepts plain text and a custom wordlist so proper nouns and technical terms aren’t false positives. LanguageTool and Hunspell both support batch runs and custom dictionaries (LanguageTool, Hunspell on GitHub).
3. Add heuristic scans to catch likely “no-space” joins and odd tokens before manual review.

Useful quick heuristics (apply to the extracted text):

• Flag very long alphabetic tokens, e.g. \b[A-Za-z]{25,}\b — often concatenations from missing spaces.
• Flag camelCase joins with a simple pattern like ([a-z])([A-Z]).
• Run a lightweight typo tool for common mistakes (e.g., codespell) to catch frequent misspellings.

Keep a project-specific ignore list and a backup before automated replacements. Automate changes only when replacements are high-confidence; otherwise produce a change list and patch the raw HTML with a script that targets the original text nodes, not the whole file, to preserve formatting. Finally, as @samarudge suggested, include an automated spell-check step in the pipeline — it’s fast and catches the bulk of issues, but always finish with a manual pass for context-sensitive fixes.