Tetragon Editorial Tools

Who this is for

This tool is aimed at publishers as a final quality-control exercise, and at production managers doing quality assurance – sometimes before an edit to get a sense of the prevailing style and editorial quality of a text in order to brief a copy-editor, and sometimes in order to brief a copy-editor. It can give a sense of what kind of editorial condition the book is in at any point in time. It can help catch mismatched quotes, names with transposed letters, numbers joined with hyphens, simple misspellings, inconsistent capitalisation, missing punctuation, and can show you whether an MS uses a British or American dictionary, and -ise or -ize endings, listing the words alphabetically with their frequency. And more.

What it is and what it isn’t

This is a mechanical pass, using filters and algorithms we’ve developed through 15 years of experience working with books. It’s aimed at mainly British and American publishers of fiction and non-fiction. It may be useful for the weird stuff – at least for a cursory glance – but will be most helpful for more “standard” texts. We probably wouldn’t use it for Finnegan’s Wake but it would be handy for a crime classic or a fantasy book. It almost always finds things after a human copyedit and proofread. It won’t try to enforce a book against a particular style-guide either, but it will show you what the prevailing style is at a glance, and whether it’s consistent. You can scroll quickly through various categories for a quick idea of the quality of the work done, and click on any findings to see it in the context of the manuscript.

How it works

It works through your manuscript line by line, cataloguing dominant styles and patterns, and flagging inconsistencies and irregularities against its own rules, and produces a report you can scroll through and act on (or dismiss, when the “error” is a deliberate choice). It is emphatically not a replacement for a copy-editor or proofreader!

“Smart services”

If explicitly requested, you can pass the manuscript through our private open-source AI model. This is something we control entirely (NOT a public AI model like GPT etc, who may decide they want to train on your author’s data). You can ask us more about how this works if you’re concerned, but we feel very strongly that we don’t trust large corporations blindly not to train on an author’s writing, given their history of copyright theft in order to train models on authors’ data.

• “Proofreading”: A “proofreading” pass is slower and costs more, but can sometimes catch things that an algorithm can’t, and that can slip past proofreaders – a missing word in a sentence (“she opened door and walked out”), an accidental homophone (“it had no affect at all”), a doubled word across a line break (“the the”), a comma splice, a parenthetical phrase that opens but never closes, or mismatched plural forms (“each of them were happy”). It can produce false positives, it can’t follow a complex style or intricate syntax, and it might over-flag deliberately unusual style. But we find it’s worth looking at from a quality point of view, in the way you would normally expect a spellcheck to have been done before publication. It can catch some real howlers – if it finds a lot at the end of production, that’s probably a sign you want another proofread! And of course, it can’t understand anything at all, it can just follow syntax and see what seems unusual from a grammatical point of view.
• Serial / Oxford comma detection: This is still a little experimental and may not be worth the time and cost unless passions run truly high (for some people, it seems, they really do), but here we try to look at all possible sentences containing lists in the manuscript and run them sequentially, flagging up which we think use serial commas and which don’t. A basic interpretation of these is given in the standard (non-smart) report, but it can’t tell which sentences are lists of things and which aren’t, so there is a lot of noise. This cuts out a lot of that, for those who care.

How you use it

Upload a .docx or .txt manuscript. The tool reads it end-to-end and produces a single self-contained HTML report. Each finding is anchored to a line number, so you can read the report alongside the manuscript in Word or InDesign and decide, item by item, whether to act on it. Italics are disregarded, at least at present (sorry). The report is read-only by design – you need to make corrections in your normal editor, whether that’s Word or InDesign, or by annotating a PDF for your typesetter.

The report

The report is divided into sections, each colour-coded and toggleable. Findings are in the sidebar; clicking one jumps you to the relevant line in the manuscript pane. Use the arrow keys liberally!

Spelling and language

• Dialect. Detects whether the manuscript is predominantly American, British -ise or British -ize, and uses that to set its expectations for everything that follows. Mixed styles are flagged side by side so it’s easy to see what’s what, and can save a lot of work if switching from US to UK English, or vice versa.
• Spelling. A dictionary check, with proper nouns and possessives recognised so they don’t generate noise. Anything unrecognised is flagged for a human to look at.
• Spelling consistency. Words that appear in two slightly different forms (one accented, one not; one transliteration vs another; close-but-not-identical proper nouns) – the kind of thing that’s easy to miss when the two forms are 200 pages apart.
• -ise / -ize and -ed / -t variants. Mixed endings within a single text (realise vs realize; burned vs burnt; amongst and among).

Punctuation and typography

• Quote marks. Tracks open and close quotes paragraph by paragraph and reports any that don’t balance (opened but not closed, for example).
• Quote punctuation. Samples how punctuation sits next to closing quotes (inside vs outside), so mixed conventions are visible at a glance. Most useful for US texts, which should have almost all punctuation inside quotes.
• Serial comma. Establishes which house style is dominant in the text, then flags lists that break it (no Oxford comma in a serial-comma book, or vice versa).
• Dash style. Categorises every em-dash, en-dash, and hyphen-as-dash so you can see at a glance whether the book is using closed em-dashes, spaced en-dashes, or a mix, and which is dominant.
• Ellipses. Spaced points vs ellipsis characters vs three full points, at a glance.
• Number ranges. Digit-hyphen-digit ranges (1939-1945) that should be en-dashes.
• Terminal punctuation. Paragraphs of five or more words that end without any closing punctuation at all – sometimes a chapter title, sometimes just a mistake.
• Special sorts. Unusual or stray characters that may be artefacts from the conversion out of Word or InDesign, also useful as part of a typesetter’s brief.

Word forms

• Hyphenation. Compounds that appear both hyphenated and unhyphenated in the same book (copy-editor vs copyeditor).
• Capitalisation. Words appearing both capitalised and lowercased mid-sentence – useful for picking up inconsistent treatment of titles, ranks, and proper-noun-adjacent terms.

References

• URLs. Every URL detected in the manuscript is checked. Each one is then visited to see whether it actually resolves: live, dead (DNS or timeout), redirected or unreachable for some other reason.
• Headings and title lines. A list of probable chapter and section headings, useful for quickly spotting an inconsistent capitalisation pattern across a contents list, or seeing the structure of a book at a glance.

Your manuscript

Manuscripts are copyrighted material. The file you upload is deleted as soon as processing finishes. The resulting report, which contains an annotated copy of the text, is kept for fourteen days by default, and you can delete it as soon as you like (though of course you should download it first if you want to keep a copy or it’ll be gone forever). Alternatively, if you want to share a report with a colleague or author, you can generate an unguessable share link (and decide when to stop sharing it). We keep just the logs of the filename, word count, when it was processed, whether it succeeded – and never the manuscript text itself once the report has expired or been deleted by you.

Access

Accounts are invite-only during early access. If you’re a publisher or editor and want to try the tool on a real title, get in touch.

Contact

You can contact us for more information at info@tetragonpublishing.com.