Skip to content

Journal Text Standardization

journal-standardize audits an existing manuscript or note against a journal corpus that was already collected with journal-analyze. It builds a vocabulary from the corpus, compares ordinary prose in the input text with that vocabulary, and writes review files for out-of-corpus wording.

Use this after you have inspected the corpus and decided that it is a suitable style reference. The tool is a writing-support audit, not an automatic acceptance filter and not a source-text copier.

Quick start

scibudy journal-analyze \
  --journal nature-communications \
  --query "atmospheric chemistry Bayesian inference" \
  --target-dir ./nc-style \
  --target-size 100

scibudy journal-standardize \
  --corpus-dir ./nc-style \
  --input ./manuscript.tex

Preview without writing output files:

scibudy journal-standardize \
  --corpus-dir ./nc-style \
  --input ./manuscript.tex \
  --dry-run

Outputs

The output directory defaults to <corpus-dir>/standardization/<input-stem>/ and contains:

  • journal_vocabulary.csv: corpus word counts and document counts.
  • journal_bigrams.csv and journal_trigrams.csv: recurring corpus word combinations.
  • allowed_terms.txt: technical terms excluded from ordinary-prose OOV checks.
  • oov_report.csv: manuscript words outside the corpus after exemptions.
  • replacement_suggestions.csv: candidate replacements for manual review.
  • standardization_summary.json: run-level counts and paths.
  • standardized.*: written only when --apply --replacement-map is used.

Applying replacements

By default, Scibudy does not rewrite the input file. To write a standardized copy, provide a CSV replacement map:

word,replacement
customword,analysis

Then run:

scibudy journal-standardize \
  --corpus-dir ./nc-style \
  --input ./manuscript.tex \
  --replacement-map ./replacements.csv \
  --apply

The original file is never overwritten. The standardized copy is written into the output directory.

LaTeX behavior

LaTeX mode is enabled by default. It skips commands, citations, labels, references, URLs, figure paths, table environments, bibliography content, and the title. Use --keep-title when the title should be audited and --plain-text for .txt or .md files that should not be treated as LaTeX.

MCP use

The same behavior is available through standardize_journal_text. Use it when Codex should audit a manuscript against a local corpus and return generated artifact paths.

Limits

  • The audit checks vocabulary presence, not scientific correctness.
  • Out-of-corpus words can be valid domain terms; add them through --allowed-term or allowed_terms.txt.
  • Replacement suggestions are candidates only and require human review.
  • Do not copy source article text; use corpus statistics and short recurring phrases only.