Kappa entry length versus translation quality

All and only the 100 headwords in Gabe's frozen final Kappa review tracker are shown. Quality is reference similarity for gpt-5.6-sol prompt v3, not an expert judgment of correctness. Select a metric, search for a headword, or click a point.

Greek length counts whitespace-delimited items containing a Greek letter, including one-letter words. Editorial brackets within a word do not split it.

Official headwords100

Mean Greek source length31.8 words

Mean four-metric quality73.1%

Length–quality Pearson r-0.494

Interactive chart

Quality metric Find headword Show only ten-entry review set

Blue circles / solid line: quotation not presentRust triangles / dashed line: quotation presentLarger points: decile-stratified review set

Select a point to see its headword, length, score, and reference-page link.

Do explicit quotations predict better translation scores?

No. In this cohort, entries with an explicit quotation scored lower, not higher. Their unadjusted four-metric mean was 62.9% versus 76.0% for entries without one.

The raw difference is statistically significant under a two-sided Welch comparison: -13.1 pp (95% CI -17.8 pp to -8.4 pp; p=1.53e-06; Cohen's d=-1.17). This is a descriptive group comparison, not evidence that quotation itself caused the lower score.

Quotation entries were also much longer: 64.7 versus 22.5 Greek words on average. After a linear adjustment for source length, the quotation association was -6.7 pp (95% CI -13.3 pp to -0.1 pp; HC3 p=0.047).

The incremental predictive value is weak. In leave-one-out cross-validation, adding the quotation flag to a linear length model changed R² from 0.191 to 0.218, while mean absolute error changed from 8.73 to 8.74 score points. Alternative log-length and quadratic-length specifications give different estimates of the incremental benefit (below). This additive model assumes parallel slopes; the separate-line analysis also examines that assumption.

Entries with explicit quotation22 / 100

Raw mean-score difference-13.1 pp

Raw Welch p1.53e-06

Length-adjusted difference-6.7 pp

LOOCV R² change+0.027

Score-by-score comparison

Reference-similarity score	Quotation mean	No-quotation mean	Raw difference	Raw 95% CI	Welch p	Length-adjusted effect	Adjusted 95% CI	HC3 p
Four-metric mean	62.9%	76.0%	-13.1 pp	-17.8 pp to -8.4 pp	1.53e-06	-6.7 pp	-13.3 pp to -0.1 pp	0.047
BLEU-4	41.2%	59.7%	-18.4 pp	-25.3 pp to -11.6 pp	1.85e-06	-9.9 pp	-19.4 pp to -0.3 pp	0.044
chrF++	66.3%	78.1%	-11.8 pp	-16.2 pp to -7.4 pp	2.88e-06	-6.4 pp	-12.7 pp to -0.0 pp	0.048
METEOR	71.5%	83.2%	-11.7 pp	-16.8 pp to -6.6 pp	4.87e-05	-5.6 pp	-12.5 pp to +1.2 pp	0.107
ROUGE-L	72.5%	82.9%	-10.4 pp	-14.6 pp to -6.2 pp	9.36e-06	-4.9 pp	-10.3 pp to +0.5 pp	0.076

Out-of-sample sensitivity to the length model

Length specification	Length-only R²	Length + quotation R²	Change	Length-only MAE	Length + quotation MAE
Linear words	0.191	0.218	+0.027	8.73%	8.74%
Log words	0.302	0.304	+0.002	7.97%	8.00%
Quadratic words	0.231	0.226	-0.005	8.37%	8.42%

Operational definition. “Explicit quotation” means that the Greek source contains at least one balanced “…” span. This reproducible rule excludes guillemets used around individual letters or forms. It does not identify uncited paraphrases or parallel passages, and it is only a proxy for possible prior representation in model training data. The outcome is entry-level similarity to the approved house translation for one gpt-5.6-sol v3 run per entry, not an expert correctness score. The primary regression is ordinary least squares with a linear word-count term and HC3 heteroskedasticity-robust uncertainty; individual metric rows are exploratory and not multiple-comparison adjusted.

Length, quotations and vocabulary

Separate ordinary least-squares lines describe entries with and without explicit quotations. Each line spans only its group’s observed source lengths. Colours, point shapes and line styles distinguish the groups.

Length versus translation similarity with separate quotation and no-quotation regression lines

Paper figure: PDF · SVG · PNG

Category	Entries	Score points per ten words	95% HC3 interval
Quotation not present	78	-3.33	-4.65 to -2.02
Quotation present	22	-0.56	-1.44 to 0.31

The quotation-minus-no-quotation slope difference was +2.77 points per ten words (95% HC3 CI 1.22 to 4.32; interaction p = 0.0006). The evidence for different slopes weakened when using log length (interaction p = 0.186) or restricting both groups to their overlapping 13–81 word range (n = 74; p = 0.393). These exploratory regressions do not establish that quotations or register changes cause difficulty.

Vocabulary associated with lower scores after allowing for length

The vocabulary-plus-length model achieved held-out R² = 0.447 and mean absolute error 7.06 score points.

The table lists the eight most negative coefficients in a ridge model combining Greek unigrams and bigrams with standardized log(1 + Greek words). Terms occur in at least two entries. The vectorizer and penalty selection were fitted inside each training fold; ten outer folds assessed prediction, and five inner folds selected the penalty by mean absolute error. Coefficients below come from a final full-cohort refit with the penalty selected by five-fold cross-validation.

Greek word or phrase	Entries	Score points per 0.1 TF–IDF	Negative / available outer fits
χωρίον	2	-1.53	9/9
δευτέρῳ	3	-1.30	10/10
πόλις ὡς	2	-1.24	8/8
οἰκήτωρ	6	-1.05	10/10
καὶ	66	-0.92	10/10
θηλυκῶς	5	-0.90	10/10
καὶ θηλυκῶς	5	-0.90	10/10
εἰ	4	-0.88	10/10

TF–IDF measures the weight of a term in an entry. These are jointly fitted, regularized associations, conditional on length and other vocabulary; they are neither independent significance tests nor effects of adding a word. The terms are accent-normalized surface forms, not lemmatized vocabulary. Fold signs describe stability across overlapping training sets, not independent replications.

Across 95 source-matched entries, a ten-percentage-point increase in the share of rare tokens was associated with +1.16 score points after adjustment for log(1 + Greek words) (95% HC3 CI -1.64 to 3.96; p = 0.412). Rare means a normalized lemma recorded fewer than 50 times in Diorisis or absent from it. This does not reproduce the strong negative rarity association reported by Zainaldin et al. for Galen; the corpora, models and outcomes differ, and reference similarity is not expert-assessed accuracy. Entries without a source-matched rarity measure were excluded from this comparison, rather than assigned a rarity of zero.

All vocabulary coefficients · Regression estimates and methods

Which source features predict translation similarity?

The best out-of-sample R² came from Greek vocabulary + log length. It explained 44.7% of held-out score variance (nested-CV R²=0.447) with mean absolute error 7.06%. Greek vocabulary + log length had the lowest MAE at 7.06% (R²=0.447).

Source structure reached R²=0.324, slightly above log length alone at 0.296, but that block includes character count, sentence count, and words per sentence and therefore carries length-adjacent signal. The standalone rarity, citation/entity, and recogniser blocks were weaker at R²=0.061, 0.067, and 0.194.

The all-available model reached R²=0.363. Removing quotation features changed R² to 0.354 and MAE from 7.72% to 7.62%. In paired bootstrap resampling, adding quotation did not clearly change error: ΔMAE +0.10 pp (95% CI -0.21 to +0.41).

The same result appears in the narrower check: adding quotation to log length changed R² from 0.296 to 0.281, with ΔMAE +0.18 pp (95% CI -0.15 to +0.48). Quotation is therefore useful as a descriptive marker of difficult, long entries, but not as a stable independent predictor in this sample.

Outer validation10-fold

Inner tuning5-fold

Best held-out R²0.447

Lowest held-out MAE7.06%

Nested cross-validated feature-block comparison

Each row predicts the same four-metric mean for unseen entries. The ridge penalty, vocabulary, document frequencies, imputation, and scaling are all learned inside the training folds. The highlighted row has the highest held-out R²; feature counts are training-fold medians.

Source feature set	Features	CV R²	CV MAE	CV RMSE	MAE gain vs mean	Spearman r	Median ridge α
Training-fold mean	0	-0.028	10.14%	12.48%	+0.00 pp	-0.326	—
Quotation only	3	0.144	9.42%	11.38%	+0.73 pp	0.134	30
Log source length	1	0.296	8.03%	10.33%	+2.11 pp	0.562	0.01
Log source length + quotation	4	0.281	8.21%	10.43%	+1.93 pp	0.557	3
Source structure	7	0.324	7.85%	10.12%	+2.29 pp	0.578	30
Lexical rarity	3	0.061	9.38%	11.92%	+0.76 pp	0.358	1
Citations + named entities	9	0.067	9.58%	11.88%	+0.56 pp	0.283	20
Guidance recogniser hits	78	0.194	8.56%	11.05%	+1.58 pp	0.467	100
Greek vocabulary	313	0.437	7.30%	9.23%	+2.84 pp	0.692	0.3
Greek vocabulary + log length	314	0.447	7.06%	9.15%	+3.08 pp	0.691	0.3
All available source features, without quotation	411	0.354	7.62%	9.89%	+2.52 pp	0.632	100
All available source features	414	0.363	7.72%	9.82%	+2.42 pp	0.635	100

What went into the all-available model?

Only information present in the Greek source before translation: log word count; explicit-quotation count and quoted-word proportion; sentence and punctuation structure; Diorisis lexical-rarity ratios; structured citation and named-entity counts; guidance-recogniser rule hits; and Greek TF-IDF unigrams and bigrams. AI-output length and any translation-similarity metric are excluded from the predictors.

Coverage. Feature source: live PostgreSQL feature tables. Recogniser hits cover 100 / 100 entries; entity features 100 / 100; citations 95 / 100; current-version rarity and sentence segmentation 97 / 100 and 99 / 100. Source-version mismatches are treated as missing and imputed within training folds. Parsed dependency-grammar coverage is 0 / 100, so it is not included. These are predictive associations in a 100-entry observational cohort; individual ridge coefficients are not treated as stable explanations.

Ten-entry review set

The 100 entries are sorted by Greek source length into ten equal groups. The lowest four-metric score in each group is selected. This gives one challenge entry from every length decile and avoids reducing review to only the ten longest or ten lowest-scoring entries.

Length decile	Headword	Greek words	Mean quality	Review text
1	Καικῖνον	6	54.5%	Open source and translations Greek: Καικῖνον, χωρίον Ἰταλικόν. Φίλιστος δευτέρῳ Σικελικῶν. gpt-5.6-sol v3: Kaikinon: a locality in Italy. Philistos, book 2 in his Sikelika. Human-approved: Kaikinon: an Italian locality. Philistos in book 2 of his Sikelika.
2	Καλάσιρις	10	63.5%	Open source and translations Greek: Καλάσιρις, μοῖρα Αἰγύπτου. Ἡρόδοτος δευτέρῳ. οἱ καὶ Καλασίριοι καὶ Καλασιρίαι. gpt-5.6-sol v3: Kalasiris: a region of Egypt. Herodotus, book 2. Its inhabitants are called both 'Kalasirioi' and 'Kalasiriai'. Human-approved: Kalasiris: a part of Egypt. Herodotos, book 2. They are 'Kalasirioi' and 'Kalasiriai'.
3	Καλαβρία	13	69.0%	Open source and translations Greek: Καλαβρία, χώρα πλησίον τῆς Ἰταλίας. ὁ οἰκήτωρ Καλαβρός, καὶ Καλαβρά θηλυκὸν καὶ Καλαβρίς. gpt-5.6-sol v3: Kalabria: a region near Italy. The inhabitant is 'Kalabros'; the feminine is 'Kalabra', and also 'Kalabris'. Human-approved: Kalabria: a region near Italy. An inhabitant is a 'Kalabros'; the feminine is 'Kalabra' as well as 'Kalabris'.
4	Κασώριον	15	51.0%	Open source and translations Greek: Κασώριον, τὸ πορνεῖον. Ἀριστοφάνης “ἐν κασωρίοισι λείχων”. ὁ οἰκήτωρ κασωρίτης, καὶ θηλυκῶς κασωρῖτις καὶ κασωρίς. gpt-5.6-sol v3: Kasorion: a brothel. Aristophanes: ‘licking in brothels’. Its patron is called a ‘kasorites’, and the feminine forms are ‘kasoritis’ and ‘kasoris’. Human-approved: Kasorion: a brothel. Aristophanes: 'licking in the kasorion...'. An inhabitant is a 'kasorites'; and in the feminine 'kasoritis' or 'kasoris'.
5	Κριώα	17	57.1%	Open source and translations Greek: Κριώα, δῆμος τῆς Ἀντιοχίδος φυλῆς. ὁ δημότης Κριωεύς. τὰ τοπικὰ Κριῶθεν κατὰ συγκοπὴν καὶ Κριῶζε καὶ Κριῶσιν. gpt-5.6-sol v3: Krioa: a deme in the tribe Antiochis. The deme-member is called 'Krioeus'. The locative terms are 'Kriothen' by syncope, 'Krioze', and 'Kriosin'. Human-approved: Krioa: a deme of the Antiochis tribe. A deme-member is a 'Krioeus'. The locative terms are 'Kriothen' (through syncope) as well as 'Krioze' and 'Kriosi'.
6	Κάληρος	25	51.8%	Open source and translations Greek: Κάληρος, οὕτως ἐκαλεῖτο ἡ Ἀλωπεκόννησος ἀπὸ Καλήρου βασιλέως, Ἀλωπεκόννησος δὲ ὅτι ἐκεῖ εἶδον σκύμνους φέρουσαν καὶ κατατιθεῖσαν ἀλώπεκα. τινὲς δ´ ὅτι δεκάληρος ἐκαλεῖτο ἡ πόλις. gpt-5.6-sol v3: Kaleros: Alopekonnesos used to be called this after King Kaleros; it was called Alopekonnesos because there they saw a fox carrying and setting down cubs. Some, however, say that the city was called 'Dekaleros'. Human-approved: Kaleros: this used to be what Alopekonesos was called, after the king Kaleros; it is now 'Alopekonesos' because they saw a fox carry her cubs there and put them down. Some say that the city used to be called 'Dekalaros'.
7	Κύρνος	34	59.9%	Open source and translations Greek: Κύρνος, νῆσος πρόσβορρος Ἰαπυγίας. Ἑκαταῖος Εὐρώπῃ. τὸ ἐθνικὸν Κύρνιοι καὶ Κυρναῖοι. φασὶ δὲ “τοὺς Κυρναίους πολυχρονιωτάτους εἶναι (οἰκοῦσι δὲ οὗτοι περὶ τὴν Σαρδόνα) διὰ τὸ μέλιτι ἀεὶ χρῆσθαι. πλεῖστον δὲ τοῦτο γίνεται παρ´ αὐτοῖς”. gpt-5.6-sol v3: Kyrnos: an island to the north of Iapygia. Hekataios in his Europe. The ethnonyms are 'Kyrnioi' and 'Kyrnaioi'. They say that ‘the Kyrnaioi are exceedingly long-lived (they dwell around Sardona) because they continually use honey; and this is produced in very great abundance among them’. Human-approved: Kyrnos: an island to the north of Iapygia. Hekataios in his Europe. The ethnonym is 'Kyrnioi' or 'Kyrnaioi'. They say 'that the Kyrnaioi are extremely long-lived (those who live round about Sardinia) because they constantly make use of honey, and this is produced by them in the greatest quantities.'
8	Κάναστρον	43	47.7%	Open source and translations Greek: Κάναστρον, ἄκρον Θρᾴκης καὶ Μακεδονίας. τὸ ἐθνικὸν Καναστραῖος. Σοφοκλῆς δὲ ὑπομνηματίζων τὰ ἀργοναυτικά “Καναστραῖον” φησίν “ἀκρωτήριον τῆς Παλλήνης”. ἀλλ´ ἐναντιοῦται τὰ ἐθνικά, εἰ μὴ καὶ τοῦτο ἐκλάβοιμεν παραπλησίως τῷ Λέχαιον καὶ Λεχαῖος, καὶ Λύκειον τὸ γυμνάσιον καὶ Λυκεῖος Ἀπόλλων, καὶ Νύμφαιον καὶ Νυμφαῖος. gpt-5.6-sol v3: Kanastron: a cape in Thrace and Makedonia. The ethnonym is 'Kanastraios'. Sophokles, commenting on the Argonautic material, says: ‘Kanastraion, a promontory of Pallene.’ But the ethnonyms are at variance, unless we understand this too in a manner similar to 'Lechaion' and 'Lechaios', 'Lykeion' the gymnasium and Apollo Lykeios, and 'Nymphaion' and 'Nymphaios'. Human-approved: Kanastron: a cape of Thrace and Macedonia. The ethnonym is 'Kanastraios', but Sophokles has 'Kanastraion' in his commentary on The Argonautika, saying it is 'a cape of Pallene'. However, the ethnonyms run contrary to this, unless we were to understand this as having the same pattern as 'Lechaion' and 'Lechaios', 'Lykeion' (the gymnasium) and Apollo 'Lykeios', or 'Nymphaion' and 'Nymphaios'.
9	Κώμη	56	49.1%	Open source and translations Greek: Κώμη, ἐν ταῖς μακραῖς ὁδοῖς μέσα χωρία ἔκτισαν πρὸς τὸ κοιμᾶσθαι νυκτὸς ἐπιγενομένης, ὅθεν καὶ ἐπικέκληται, ὡς Φιλόξενος. ὁ οἰκήτωρ ἐγκώμιος. Ἡσίοδος “εἰ γάρ τοι καὶ χρῆμ´ ἐγκώμιον ἄλλο γένοιτο”. καὶ κωμαῖος καὶ κωμήτης ἀρσενικῶς καὶ θηλυκῶς διὰ τοῦ « ι ». καὶ ὁ δῆμος κώμη. εἴρηται ἀπὸ τοῦ ἐν αὐταῖς κοιμᾶσθαί τε καὶ αὐλίζεσθαι. λέγεται καὶ κωμήτωρ. gpt-5.6-sol v3: Kome: on long roads they built halfway points for sleeping when night came, from which it also received its name, as per Philoxenos. The inhabitant is an 'enkomios'. Hesiod: ‘For even if some other need should arise for you in the village.’ Also 'komaios', and 'kometes' in the masculine and 'kometis' in the feminine, with ι. A deme too is a 'kome'. The term is said to derive from sleeping and lodging in them. The inhabitant is also called a 'kometor'. Human-approved: Kome: along long routes they established halfway points for the purpose of sleeping when night fell, after which they are now named, as per Philoxenos. An inhabitant is an 'enkomios'. Hesiod: 'For if some matter were to arise for you as an enkomios'. And also a 'komaios', and 'kometes' in the masculine, which in the feminine is written with ι. A deme is also a kome, named after the practice of taking shelter and sleeping there. 'Kometor' is also said.
10	Καπετώλιον	95	52.6%	Open source and translations Greek: Καπετώλιον, ἐν Ῥώμῃ λόφος Ταρπαῖος πάλαι λεγόμενος. περὶ δὲ τῆς γραφῆς τοῦ τεμενικοῦ ὁ Εὐδαίμων διὰ τοῦ « ι » φησίν “Εἰρηναῖος προπαροξύνει διὰ τοῦ « ι » γράφων”. Ἡρωδιανὸς δὲ Καπετωλίειον ἐν ἓξ συλλαβαῖς, καὶ ἐν συναιρέσει Καπετώλειον. τὸ δὲ Καπετωλεῖα ἐπὶ τοῦ ἀγῶνος προπερισπᾶται. ὅσα γὰρ ἔχει προϋπάρχοντα εἰς « ος » καθαρόν, παραληγόμενα ἢ μόνῳ τῷ « ι » ἢ προηγουμένου αὐτοῦ τοῦ « α » ὥστε εἶναι πρὸ τέλους τὴν « αι » δίφθογγον, προπερισπᾶται, ἢ καὶ ὅσα κτητικά. Ἀσκληπιεῖον, Ἀσκληπιός γάρ· Πτολεμαεῖον, Πτολεμαῖος γάρ· Ὀλυμπιεῖα τὰ Ἀθήνησιν, Ὀλύμπιος γάρ. τὸ δὲ τέμενος Καπετώλιον καὶ Καπετώλιος Ζεὺς διὰ τοῦ « ι ». τὸ ἐθνικὸν Καπετωλῖνος. gpt-5.6-sol v3: Kapetolion: a hill in Rome formerly called Tarpeios. On the spelling of the sanctuary name, Eudaimon says that it is written with ι: ‘Eirenaios accents it with an acute on the antepenult, writing it with ι.’ Herodianos gives Καπετωλίειον in six syllables, and, with vowel contraction, Καπετώλειον. Καπετωλεῖα, when it refers to the contest, is accented with a circumflex on the penult. For all forms that have an underlying form ending in pure -os and have in the penult either ι alone or ι preceded by α, so that the diphthong αι stands before the final syllable, are accented with a circumflex on the penult, as are possessive forms: ‘Asklepieion’, since the underlying form is ‘Asklepios’; ‘Ptolemaeion’, since it is ‘Ptolemaios’; the ‘Olympieia’ at Athens, since it is ‘Olympios’. The sanctuary is ‘Kapetolion’, and Zeus Kapetolios is written with ι. The ethnonym is ‘Kapetolinos’. Human-approved: Kapetolion: in Rome, a hill that was long ago called 'Tarpaios'. Regarding the spelling of the sanctuary name, Eudaimon says that it is with ι: 'Eirenaios accents it with acute on the antepenult and writes it with ι.' Herodian has 'Kapetolieion' with six syllables, and 'Kapetolion' with contraction. The form 'Kapetoleia', referring to the games, is accented with circumflex on the penult. This is because forms whose base already ends in postvocalic -ος—when either a single ι is in the penultimate position or α precedes it so that the diphthong αι stands before the ultima—will be accented with a circumflex on the penult, and the same applies to possessive forms. Asklepieion (Ἀσκληπιεῖον) is thus from 'Asklepios' (Ἀσκληπιός); Ptolemaeion (Πτολεμαεῖον) is thus from 'Ptolemaios' (Πτολεμαῖος); Olympieia (Ὀλυμπιεῖα), the one at Athens, is thus from 'Olympios' (Ὀλύμπιος). The sanctuary, however, is 'Kapetolion' (Καπετώλιον), and 'Zeus Kapetolios' (Καπετώλιος), with ι. The ethnonym is 'Kapetolinos'.

Downloads

Verify all 100 official headwords

Official order	Kappa entry	Headword	Greek words	Mean quality	Review set
1	1	Καβαλίς	47	80.0%
2	2	Καβασσός	74	58.0%
3	3	Καβειρία	81	62.8%
4	4	Καβελλιών	23	72.0%
5	5	Καβύλη	16	89.1%
6	6	Καδμεία πόλις	17	62.5%
7	7	Κάδοι	13	74.0%
8	8	Καδούσιοι	11	100.0%
9	9	Κάδρεμα	14	73.1%
10	10	Κάθαια	17	58.0%
11	11	Καικῖνον	6	54.5%	Decile 1
12	12	Καινίνη	10	78.0%
13	13	Καινοί	7	83.8%
14	15	Καινύς	16	72.5%
15	14	Καιρή	11	80.7%
16	16	Καισάρεια	30	78.3%
17	17	Καλαβρία	13	69.0%	Decile 3
18	18	Καλάθη	32	80.0%
19	19	Καλάμαι	5	93.7%
20	20	Καλαμένθη	16	63.5%
21	21	Κάλαρνα	10	86.8%
22	22	Καλάσιρις	10	63.5%	Decile 2
23	23	Καλατίαι	5	89.4%
24	24	Καλαύρεια	20	70.6%
25	25	Κάλβιος	15	73.3%
26	26	Καλὴ ἀκτή	49	62.2%
27	27	Κάληρος	25	51.8%	Decile 6
28	28	Καλησία	12	94.5%
29	29	Καλλάτηβος	8	100.0%
30	30	Κάλλατις	47	57.2%
31	31	Καλλίαι	17	73.4%
32	32	Καλλίαρος	31	80.3%
33	33	Καλλιόπη	11	90.2%
34	34	Καλλίπολις	31	69.0%
35	35	Κάλπη	36	77.0%
36	36	Καλύβη	14	88.6%
37	37	Κάλυδνα	29	82.1%
38	38	Καλυδών	15	82.3%
39	39	Κάλυμνα	26	85.0%
40	40	Κάλυνδα	10	85.4%
41	41	Κάλυτις	17	62.5%
42	48	Κάμιρος	47	66.6%
43	51	Κάναθα	17	68.9%
44	52	Κάναι	60	63.1%
45	53	Κάναστρον	43	47.7%	Decile 8
46	63	Κάνωπος	62	69.4%
47	66	Καπετώλιον	95	52.6%	Decile 10
48	68	Καππαδοκία	59	63.3%
49	75	Καρδαμύλη	34	78.2%
50	82	Καρία	194	57.6%
51	97	Καρπασία	83	63.9%
52	99	Καρπήσιοι	7	83.3%
53	102	Καρύανδα	25	74.7%
54	103	Κάρυστος	138	64.1%
55	104	Καρχηδών	90	55.6%
56	105	Κάσιον	45	74.0%
57	107	Κάσος	47	72.6%
58	109	Κοσσαῖος γενεὴν Κασπειρόθεν	90	57.4%
59	120	Κάσταξ	15	77.0%
60	121	Κάστνιον	20	67.5%
61	122	Καστωλοῦ πεδίον	29	74.6%
62	123	Κασώριον	15	51.0%	Decile 4
63	124	Κατάβαθμος	16	89.5%
64	125	Κατακεκαυμένη	34	69.5%
65	127	Κατάνειρα	12	66.1%
66	126	Κατάνη	68	59.6%
67	129	Καταννοί	8	88.3%
68	128	Καταονία	17	73.2%
69	130	Κάτρη	18	74.7%
70	151	Κεκρυφάλεια	17	77.5%
71	180	Κορώνεια	77	65.7%
72	188	Κοτιάειον	48	63.8%
73	205	Κρανίδες	11	96.0%
74	222	Κριώα	17	57.1%	Decile 5
75	286	Κύρβη	8	100.0%
76	287	Κύρη	15	84.8%
77	288	Κυρήνη	35	75.3%
78	289	Κύρης	10	70.7%
79	290	Κύρις	13	79.0%
80	291	Κύρνος	34	59.9%	Decile 7
81	292	Κύρου πόλις	24	66.5%
82	293	Κύρρος	33	75.0%
83	294	Κυρταία	29	73.0%
84	295	Κύρτος	49	65.3%
85	296	Κύρτωνες	19	65.7%
86	297	Κυρτώνιος	14	78.9%
87	298	Κύτα	59	71.0%
88	299	Κυτέριον	18	63.8%
89	300	Κύτινα	22	96.6%
90	301	Κυτώνιον	14	91.7%
91	302	Κύτωρος	22	74.6%
92	303	Κύφος	41	62.1%
93	304	Κυχρεῖος πάγος	47	70.2%
94	305	Κύψελα	23	84.4%
95	306	Κῶβρυς	12	86.5%
96	307	Κώθων	15	94.3%
97	308	Κωλιάς	43	67.9%
98	309	Κῶλοι	18	67.7%
99	310	Κώμη	56	49.1%	Decile 9
100	311	Κωνώπη	47	63.1%

Generated 2026-09-20 11:24:59 UTC. The four-metric mean is the unweighted entry-level mean of BLEU-4, chrF++, METEOR, and ROUGE-L against the approved house-style translation.