Early Modern English and the biblical register (`en-x-emode`)

A reference manual of the English of about 1500–1700 and of the biblical register that continues it — the King James Bible (1611), the JPS 1917 Tanakh, Shakespeare, and the old voice of 19th-century fairy tales — written for Mova: what each feature is, how it differs from modern English, how Mova parses it today, where it fails (measured), and what to change.

The tag en-x-emode is a private-use BCP 47 subtag: English, variety «Early Modern».

Chapters

File	Contents
1. The period, the register and the texts Mova reads	dates and stages; the two registers; the line of Bibles from Tyndale to JPS 1917; a measured KJV / JPS comparison (JPS keeps thou, -eth, shalt, but writes spoke, begot, its); the corpora and how examples are cited
2. Spelling, print conventions and tokens	u/v, i/j, long s, ye = the; what is left in JPS (to-morrow, for ever, LORD, divine capitals); quotation and verse marks; elision ('tis, may'st); normalisation and when it helps
3. Pronouns	thou/thee/thy/thine/thyself, ye/you; T/V use in Shakespeare vs number use in the Bible; mine/thine before vowels; his for its; relatives and indefinites; lemma and FEATS decisions; what the small model does
4. Verb endings: -est, -st, -eth, -th	-est/-st, -eth/-th, plural -s/-th/-en; the 2nd person past; spelling rules; `morph::archaic` measured on 1 484 db tokens: three bugs and their fixes
5. Auxiliaries and modals	shalt, wilt, hath, hast, doth, dost, didst, art, wast, wert, canst, mayest; meanings (shall = law, will = want, may = purpose); be-perfect (is come); UD analysis; the gate
6. Strong verbs, old pasts and participles, old plurals	spake, brake, bare, sware, wrought, smote, sod, digged, builded, holpen, quoth…; old plurals kine, brethren; what the lemmatizer returns for each
7. Do, its absence, and negation	I know not, Fear not, Knowest thou, did eat; the history of do-support; where Mova attaches postverbal not (72 errors)
8. Word order and phrase structure	inversion after adverbs and fronted objects; dislocation; appositions (the LORD thy God); partitive of; from off; unto; ethical datives
9. The subjunctive, conditions and wishes	if he be, lest ye die, until he come, Be it, cursed be; an = if; how UD and the db mark the mood
10. Vocabulary: archaic words and false friends	pro-adverbs (thereof, wherewith, whither…), discourse words (behold, peradventure…), ~50 false friends (even = evening, corn = grain, suffer = allow…), archaic nouns and verbs, with examples and Mova's tags
11. The biblical style	and-chains; it came to pass (five analyses in the db); speech framing; Hebraisms as idioms; numbers; the register in the tales and the Sefer Yetzirah
12. How Mova reads this English today: pipeline, fixes, failures	the Torah cycle stage by stage; the 17.10 fixes; refusals per round; the small model's errors by class; how wrong lemmas and features survive; the db audit (D1–D15); effects on pragmatics and the world
13. Concrete improvements for Mova's parser and lexicon	30 concrete changes (tables T1–T5, lemmatizer L1–L6, features F1–F3, syntax S1–S7, pragmatics P1–P4, world W1–W4, tokenizer K1–K2, training R1–R2), each with its count and negative control
14. Sample texts, read word by word	Genesis 1:1–5 in KJV and JPS; Genesis 3 and 2:17 word by word (db vs small model); a Grimm tale; Shakespeare lines from the grammars; Westcott's Sefer Yetzirah
Sources	grammars, papers, texts, Mova's data and code, web pages, and the «wanted» list

Quick reference

	Present	Past
thou + lexical verb	thou lovest, thou seest, thou goest	thou lovedst, thou camest, thou saidst
he + lexical verb	he loveth, he goeth, he saith	he loved, he came
be	I am, thou art, he is; subj. be	I was, thou wast/wert, he was; subj. were
have	thou hast, he hath	thou hadst
do	thou dost, he doth (aux); doest, doeth (lexical)	thou didst
modals	thou shalt, wilt, canst, mayest	thou shouldest, wouldest, couldest, mightest

Pronoun	Nom	Acc	Gen
2 sg	thou	thee	thy / thine (before a vowel, and predicative)
2 pl	ye	you	your / yours

Construction	Old	Modern
negation	I know not	I do not know
question	Knowest thou?	Do you know?
negative imperative	Fear not	Do not fear
affirmative do	he did eat (= ate)	he ate
subjunctive	if he be, lest ye die	if he is, so that you do not die
perfect of motion	they are come	they have come
condition	an thou knowest	if you know

The state on 2026-10-17, in five numbers

1 087 distinct Torah sentences were refused at the annotation gate during the day because an archaic auxiliary kept its form as its lemma (shalt 969 refusals, hath 401, hast 167, art 141); after the two fixes of the evening (3e639a08e, d3101c072) the refusal rate fell from 28 % to 0.4 % of drafts.
The small model en-agreed.bin agrees with the Torah db on 94.25 % of UPOS; archaic tokens are 6 % of the text and 41 % of the disagreements; -eth and 2nd person -est verbs are wrong 90 % of the time.
2 242 of 6 013 Torah sentences have a tagging error that a closed table and an ending rule would fix (chapter 13, T1–T3).
The db carries 400 verb tokens whose lemma is still the archaic form (giveth, goeth, hath) in 328 sentences, 98 sentences with a third-person thou, and 202 with thee lemmatized as thee — all repairable by deterministic rules (chapter 13, L5, L6, F1).
72 postverbal not are attached to the wrong word, which flips the polarity of the root in the pragmatic layer (chapter 7.4, rule S1).

How the numbers were made

Paths: $MOVA_DATA is Mova's data directory (corpora, runs, models; outside the repo), $RESEARCH_RAW the raw folder of the shared research library (papers and public-domain grammars, one folder per work with a SOURCES.md), $UD_ROOT a checkout of Universal Dependencies 2.18.

Counts over text are whole-word matches over torah.tsv (6 105 sentences, 156 603 words), the KJV Torah (eBible 1769 text, brackets removed) and the tales pool. Counts over trees are over the 6 065 JPS sentences of current.conllu (2026-10-17 21:29). «SLM now» is en annotate $MOVA_DATA/runs/mmm-parse-2026-10-15/en-agreed.bin built from main at f3e616747, run twice with byte-identical output, aligned with the db by sentence id and token index where the forms agree (6 013 sentences, 180 871 tokens). The lemmatizer probe in chapters 4 and 6 is a throw-away Rust binary that links the en crate and calls en::morph::archaic and Morph::lemmatize; it is not part of the repo. The analysis scripts were one-off counts; any number here can be reproduced from the files named in the chapter.

Scope and limits

The manual covers the biblical register thoroughly (that is what Mova reads) and living EModE (Shakespeare, letters, EEBO) from the grammars only; none of the plays is in Mova's corpora yet.
Pronunciation (the Great Vowel Shift) is out of scope: Mova reads text.
Dialect forms in the tales (wuz, kin, twa, maun) belong to other varieties.
Nothing in this folder changes code. Chapter 13 is a list of proposals for separate, measured commits.

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/en-x-emode/