Dataset Viewer
Auto-converted to Parquet Duplicate
aramaic_strong
stringlengths
1
4
aramaic_lemma
stringlengths
3
16
tokens
stringclasses
39 values
hebrew_strong
stringlengths
0
4
hebrew_lemma
stringlengths
0
16
basis
stringclasses
5 values
greek_check
stringclasses
3 values
domain_check
stringclasses
3 values
4430
מֶלֶךְ
180
4428
מֶלֶךְ
strongs
1
1
426
אֱלָהּ
95
433
אֱלֹוהַּ
strongs
1
1
3809
לָא
82
3808
לֹא
strongs
1
3606
כֹּל
82
3605
כֹּל
strongs
1
1934
הוה
71
1933
הָוָה
strongs
1
560
אֲמַר
71
559
אָמַר
strongs
1
1
1836
דְּנָה
58
2177
זַן
consonants
1
116
אֱדַיִן
57
4437
מַלְכוּ
57
4438
מַלְכוּת
strongs
1
1
3046
יְדַע
47
3045
יָדַע
strongs
1
1
1005
בַּיִת
44
1004
בַּיִת
strongs
1
1
8065
שָׁמַיִן
38
8064
שָׁמַיִם
strongs
1
6966
קוּם
35
6965
קוּם
strongs
1
1
2370
חֲזָה
31
2372
חָזָה
strongs
1
1
6591
פְּשַׁר
31
6622
פָּתַר
strongs_root
0
1
2942
טְעֵם
30
2938
טָעַם
strongs_root
0
1
6032
עֲנָה
30
6030
עָנָה
strongs
1
1
6903
קֳבֵל
29
6905
קָבָלְ־
strongs
5648
עֲבַד
28
5647
עָבַד
strongs
1
3052
יְהַב
28
3051
יָהַב
strongs
1
1
7761
שׂים
26
7760
שׂים
strongs
1
1
606
אֱנָשׁ
25
582
אֱנוֹשׁ
strongs
1
1
4406
מִלָּה
24
4405
מִלָּה
strongs
1
1
7229
רַב
23
7227
רַב
strongs
1
1
1722
דְּהַב
23
2091
זָהָב
strongs
1
1
1124
בְּנָה
22
1129
בָּנָה
strongs
1
1
2493
חֵלֶם
22
2492
חָלַם
strongs_root
1
1
1400
גְּבַר
21
1397
גֶּבֶר
consonants
1
1
772
אֲרַע
21
776
אֶרֶץ
strongs
1
1
8133
שׁנה
21
8132
שׁנה
strongs
0
1
5957
עָלַם
20
5769
עוֹלָם
strongs
1
0
6523
פַּרְזֶל
20
1270
בַּרְזֶל
strongs
1
1
2423
חֵיוָה
20
2421
חיה
strongs_root
1
0
7912
שְׁכַח
18
7911
שָׁכַח
strongs_shifted
0
0
383
אִיתַי
17
3426
יֵשׁ
strongs
1
3028
יַד
17
3027
יָד
strongs
1
1
6755
צְלֵם
17
6754
צֶלֶם
strongs
1
1
5135
נוּר
17
5216
נֵר
strongs_root
1
0
858
אֲתָה
16
857
אָתָה
strongs
0
1
5972
עַם
15
5971
עַם
strongs
1
1
2324
חֲוָה
15
2331
חָוָה
strongs
1
1
7972
שְׁלַח
14
7971
שָׁלַח
strongs
1
1
3118
יוֹם
14
3117
יוֹם
strongs
1
1
479
אִלֵּךְ
14
411
אֵל
strongs_root
0
1
7217
רֵאשׁ
14
7218
רֹאשׁ
strongs
1
1
1882
דָּת
14
1881
דָּת
strongs
1
1
2445
חַכִּים
14
2449
חָכַם
strongs_root
1
0
5954
עֲלַל
14
5953
עָלַל
strongs_shifted
0
0
7162
קֶרֶן
14
7161
קֶרֶן
strongs
1
1
7985
שׇׁלְטָן
14
7980
שָׁלַט
strongs_root
0
1
3705
כְּעַן
13
3651
כֵּן
strongs_root
1
0
1791
דֵּךְ
13
2088
זֶה
strongs_root
1
1965
הֵיכַל
13
1964
הֵיכָל
strongs
1
1
1459
גֹּוא
13
1460
גֵּו
strongs
0
0
7690
שַׂגִּיא
13
7689
שַׂגִּיא
strongs
1
1
3702
כְּסַף
13
3701
כֶּסֶף
strongs
1
1
5732
עִדָּן
13
5708
עִדָּה
strongs_root
0
6922
קַדִּישׁ
13
6918
קָדוֹשׁ
strongs
1
1
7606
שְׁאָר
12
7605
שְׁאָר
strongs
1
8036
שֻׁם
12
8034
שֵׁם
strongs
1
1
3792
כְּתָב
12
3791
כְּתָב
strongs
1
1
7412
רמה
12
7411
רָמָה
strongs
1
1
3202
יְכִל
12
3201
יָכֹל
strongs_root
1
1
1156
בעה
12
1158
בָּעָה
strongs
1
1
2376
חֱזוּ
12
2372
חָזָה
strongs_root
0
1
5457
סְגִד
12
5456
סָגַד
strongs
1
4083
מְדִינָה
11
4082
מְדִינָה
strongs
1
1
7123
קְרָא
11
7121
קָרָא
strongs
1
1
2166
זְמַן
11
2165
זְמָן
strongs_root
1
1
5312
נְפַק
11
1123
בַּר
11
1121
בֵּן
strongs
1
1
5308
נְפַל
11
5307
נָפַל
strongs
1
1
927
בְּהַל
11
926
בָּהַל
strongs
1
1
7308
רוּחַ
11
7307
רוּחַ
strongs
1
1
3062
יְהוּדַי
10
3064
יְהוּדִי
consonants
1
1
7990
שַׁלִּיט
10
7989
שַׁלִּיט
strongs
0
1
6347
פֶּחָה
10
6346
פֶּחָה
strongs
0
1
6399
פְּלַח
10
6398
פָּלַח
strongs
0
0
1855
דְּקַק
10
1854
דָּקַק
strongs
1
861
אַתּוּן
10
784
אֵשׁ
strongs_root
0
0
5943
עִלַּי
10
5942
עִלִּי
strongs
0
6634
צבה
10
6623
פִּתְרוֹן
strongs_shifted
0
0
1358
גֹּב
10
744
אַרְיֵה
10
738
אַרְיֵה
strongs
1
1
1541
גלה
9
1540
גָּלָה
strongs
1
1
7149
קִרְיָה
9
7151
קִרְיָה
strongs
1
1
2
אַב
9
1
אָב
strongs
1
1
7127
קְרֵב
9
7126
קָרַב
strongs
1
1
7328
רָז
9
5709
עֲדָה
9
5710
עָדָה
strongs
0
0
5174
נְחָשׁ
9
5154
נְחוּשָׁה
strongs
1
1
2635
חֲסַף
9
2636
חַסְפַּס
strongs_root
0
324
אֲחַשְׁדַּרְפַּן
9
323
אֲחַשְׁדַּרְפַּן
strongs
1
1
8086
שְׁמַע
9
8085
שָׁמַע
strongs
1
7804
יזב
9
5800
עָזַב
strongs
0
0
3790
כְּתַב
8
3789
כָּתַב
strongs
1
1
524
אֻמָּה
8
523
אֻמָּה
strongs
0
1
5559
סְלֵק
8
5559
סָלֵק
consonants
1
1
1247
בַּר
8
1121
בֵּן
strongs
1
1
3652
כֵּן
8
3651
כֵּן
strongs
1
End of preview. Expand in Data Studio

Aramaic to Hebrew equivalents (v0.1.1)

The Aramaic parts of the Tanakh (Daniel 2:4b-7:28, Ezra 4:8-6:18 and 7:12-26, Jeremiah 10:11, two words in Genesis 31:47) use their own lexemes, with their own Strong's numbers. Whatever is known of a Hebrew word (its renderings in translations, the Septuagint's Greek for it, a lexicon entry) does not reach its Aramaic counterpart. Above all, a Hebrew-Greek lexeme bridge (which Greek lexemes render a Hebrew one, as used to align words across translations) has no entry for an Aramaic word, so an aligner gets no Greek candidates for it. Two tables make that link, by rule, from the original-language sources only: through them an Aramaic word takes the bridge entries, and anything else keyed by Hebrew lexemes, of its Hebrew counterparts.

lexemes: the Hebrew cognate of each Aramaic lexeme (600 content lexemes, proper names left out)

column description
aramaic_strong, aramaic_lemma, tokens the Aramaic lexeme (MACULA Hebrew, wlang A) and its content-word occurrences
hebrew_strong, hebrew_lemma its Hebrew cognate
basis strongs: Strong's dictionary says "(Aramaic) corresponding to" that Hebrew number; strongs_shifted: the same, "in the sense of" or "through the idea of" another meaning (same root, meaning moved: Aramaic שְׁכַח "find", Hebrew שָׁכַח "forget"; not for word choice); strongs_root: Strong's derives it "from" that root; consonants: no Strong's link, the same consonants after the regular shifts ד/ז, ת/שׁ, ט/צ, ע/צ, chosen by Septuagint overlap; empty: none
greek_check, domain_check 1 / 0: whether the two share Septuagint Greek, and a UBS (SDBH) domain code; empty where either lacks one
basis rows
strongs 362
strongs_root 143
(none) 75
consonants 10
strongs_shifted 10

Only Strong's cross-reference numbers are read, never its English.

words: Hebrew words of the same meaning, per Aramaic word (3,278 words)

A cognate does not always mean the same: Aramaic שְׁכַח is "find", Hebrew שָׁכַח "forget"; Aramaic פְּלַח "serve" matches Hebrew עָבַד, not its cognate; many words (Persian titles, נוּר "fire", עִדָּן "time") have no Hebrew cognate in use. So each Aramaic content word gets a ranked list of Hebrew words that can say the same thing at that place, synonyms and related words, of any length: one word where one stands out, many where several compete. Their bridge entries together give an aligner the Greek (and, through it, the translations' words) to look for.

  • Candidates: Hebrew content lexemes of the same word class (noun and adjective counted together), with 2+ occurrences, that carry one of the word's UBS (SDBH) domains on at least 5 % of their own occurrences; and the cognate.
  • Score: how typical that meaning is for the candidate (the share of its occurrences carrying the word's most specific domain codes, plus 0.3 x the share for any of its codes), times how often the Septuagint renders the candidate by the Greek at the place (MACULA's Greek link for the word, else the Greek verse: Theodotion for Daniel, GLAUx for Ezra), at least 0.01; the cognate's score x 1.5.
  • The list: every candidate scoring at least 0.2 x the best, best first; a safety cap at 15.
  • The pick (hebrew_strong): one word, for a rule engine. The cognate is kept where at least 30 % of its own occurrences carry the word's domain; otherwise the best equivalent with 10+ occurrences, carrying the domain on 20 % of its own.
column description
key the MACULA Hebrew word id
aramaic_strong, aramaic_lemma the Aramaic lexeme
hebrew_strong, hebrew_lemma, basis the pick: cognate or equivalent
evidence cognate_meaning, greek_word, greek_verse or domain: what decided the pick
score, margin the pick's score, and its ratio to the next
cognate_strong the cognate (lexemes, basis strongs or consonants)
candidates, candidate_scores the ranked list (Hebrew Strong's numbers, space-separated, the pick first) and their scores

Lists on the Aramaic words (3,191 with one): mean length 5.1; a single word 0.38; ten or more 0.24. A long list means the evidence does not settle between the candidates, not that all are equally good.

How good it is

  • Where both lexemes have UBS domains, the plain strongs links share one 0.86 of the time, strongs_root links only 0.41: use those as a fallback. The Greek check is weaker evidence, since the Old Greek of Daniel 3-6 is free.
  • In drafting (a rule-based engine drafting Daniel's 196 Aramaic verses with Daniel held out of its training; chrF3), with the strongs and consonants links: Indonesian 27.96 -> 31.74, Swahili 25.78 -> 28.89, Tagalog 25.69 -> 28.78; Daniel's Hebrew verses unchanged. Adding strongs_root: a further +0.4 to +1.0. The share of Aramaic content words that get any word rises from 0.64 to 0.82. Exact word choice rises much less: the Hebrew cognate's usual rendering is often not the word a translation uses for the Aramaic.
  • The words lists, checked on Hebrew (the same procedure on the Hebrew content words of Daniel and Ezra, each word hidden and no cognate given, so the lists must find the word itself): the word is on its list 0.723 of the time, lists averaging 3.3 words, a single word 0.45 of the time; the single pick alone finds it 0.516. A near-synonym counts as a miss, so these are lower bounds. With the word standing in for its cognate, the pick finds the word itself 0.899 of the time (2,941 words).
  • For the bridge (1,519 Aramaic words whose own Greek link MACULA records; that link hidden, the lists ranked from the Greek verse, the information an aligner has): the hidden Greek is among the bridge's Greek for the cognate alone 0.500 of the time; for the list with the cognate 0.648 (a mean 3.5 Hebrew lexemes offered per word); for the single pick 0.533.
  • Coverage of the translations' words (the same engine and test; for each Aramaic word linked to a word in a translation, is that word among the renderings the engine learned for the Hebrew side): cognate alone / single pick / list, Indonesian 0.330 / 0.366 / 0.477 (2,182 words), Swahili 0.230 / 0.266 / 0.356 (2,045), Tagalog 0.277 / 0.303 / 0.400 (2,085): the list reaches +12 to +15 points more. Long lists add too: words with 10+ candidates gain about as much per word as words with 2-9. A list consults about 5 Hebrew lexemes and 90-150 renderings per word, against 1 and 35-60 for the cognate, so this measures reach, not precision.
  • Voting over the list (every listed lexeme votes for its renderings, weighted by score) made the drafts worse: exact words -2 to -5 points, as frequent lexemes' general renderings outweigh the cognate's specific one. Choosing from a list needs context (an aligner's own evidence at the place), not a flat vote.
  • The pick in drafting (the same engine and test, the pick in place of the cognate): exact content words +0.0 to +0.5 points and chrF3 +0.3 to +0.7, small. A rule engine that takes one word gains little from it; the lists serve the bridge and alignment.
  • Known weak spots: some time words (night, season) rank "month" high; officials' titles map to the nearest Hebrew offices. No person has checked the tables.

Licence

CC BY-SA 4.0. Built from MACULA Hebrew (CC BY 4.0), the UBS dictionary's senses and domains (CC BY-SA 4.0), Strong's Hebrew dictionary (1890, public domain; OpenScriptures XML, CC BY 4.0), MACULA's Hebrew-Greek links and the Septuagint from GLAUx (CC BY-SA 4.0; Theodotion's Daniel, document 0527-057). Credit: MACULA (Clear Bible / Biblica), United Bible Societies, OpenScriptures, GLAUx (KU Leuven), make-tanakh-research (the drafting measurements), and this dataset.

Downloads last month
79