NLP ܒܠܫܢܐ ܥܒܪܝܐ: ܡܐܢ̈ܐ ܘܐܘܠܨܢ̈ܐ

ܡܕܒܪܢܐ (guide)

ܡܐ ܐܝܬ ܓܘ ܐܗܐ ܦܐܬܐ (On this page)

ܦܘܠܚܢܐ ܕܠܫܢܐ ܛܒܝܥܝܐ (natural language processing, NLP) ܝܠܗ ܬܚܬܝܬܐ ܕܬܚܘܬ ܟܠ ܦܘܫܩܐ ܟܡܝܝܐ ܕܡ̈ܠܐ – ܡܢܝܢܐ ܕܡ̈ܠܐ، ܦܘܫܩܐ ܕܬܚܘܡܐ ܕܡ̈ܠܐ (content analysis)، topic modeling، ܡܫܬܘܕܥܢܘܬܐ ܕܪܓܫܐ. ܓܘ ܐܢܓܠܝܙܝܐ ܐܗܐ ܬܚܬܝܬܐ ܫܦܝܬܐ ܝܠܗ̇: ܟܦܪܫܚ ܒܦܬܝ̈ܘܬܐ، ܟܫܩܠܚ ܫܘܠܡܐ ܕܡܠܬܐ، ܟܡܢܚ. ܓܘ ܥܒܪܝܐ ܐܝܗܝ̇ ܠܐ ܝܠܗ̇ ܫܦܝܬܐ، ܘܒܨܘܝܐ ܕܚܫܒ ܠܗ̇ ܐܝܟ ܫܦܝܬܐ ܒܕ ܩܒܠ ܡܢܝ̈ܢܐ – ܛܥܝ̈ܐ ܒܫܠܝܐ. ܦܐܬܐ ܟܡܦܫܩܐ ܡܘܕܝ ܓܘ ܥܒܪܝܐ ܟܬܒܪ ܡܐܢ̈ܐ ܟܠܢ̈ܝܐ (ܦܣܩܐ 1)، ܐܘܠܨܢܐ ܐܘܡܢܝܐ ܕܩܕܡ ܟܠ ܦܘܫܩܐ – ܪܡܙܢܘܬܐ (encoding) ܘܐܘܪܚܐ ܕܟܬܝܒܬܐ (2)، ܡܘܕܝ ܟܥܒܕ ܨܢܘܪܐ (pipeline) ܬܪܝܨܐ ܕܦܘܠܚܢܐ ܕܥܒܪܝܐ ܘܐܝܢܝ ܡܐܢ̈ܐ ܟܦܠܚܝ ܠܗ (3)، ܐܝܢܝ ܝܢܐ ܡܘܕܝ̈ܠܐ ܕܠܫܢܐ ܕܐܝܬ (4)، ܐܝܢܝ corpora ܘ-treebanks ܐܝܬ (5)، ܡܘܕܝ ܦܪܝܫܐ ܝܠܗ ܓܘ ܥܒܪܝܐ ܕܟܬܒܐ ܩܕܝܫܐ ܘܕܪ̈ܒܢܐ (6)، ܘܛܘܦܣܐ ܕܥܒܕܐ ܥܒܕܢܝܐ ܩܐ ܦܘܫܩܐ ܕܡ̈ܠܐ ܥܒܪ̈ܝܐ ܓܘ ܒܘܨܝܐ ܟܢܫܝܐ (7) – ܗܠ ܢܘܩܙܐ ܕܓܘܗ ܡ̈ܠܐ ܟܗܘܝܢ ܡܢܝ̈ܢܐ، ܘܡܚܫܒ̈ܢܐ ܕܣܐܝܛ ܟܡܫܪܝ.

1. ܩܐ ܡܘܕܝ ܥܒܪܝܐ ܥܣܩܬܐ ܝܠܗ̇ ܩܐ ܚܘܫܒܢܐ – ܐܪܒܥܐ ܐܘܠܨܢ̈ܐ

ܐ. ܟܬܝܒܬܐ ܕܠܐ ܢܘ̈ܩܙܐ ܝܠܗ̇، ܘܡܛܠ ܗܕܐ ܥܡ ܬܪܝܢܘܬܐ ܕܣܘܟܠܐ. token ספר ܡܨܐ ܕܗܘܝܐ סֵפֶר، סַפָּר، סָפַר، סִפֵּר ܝܢ סְפָר – ܚܡܫܐ ܡ̈ܠܐ ܦܪ̈ܝܫܬܐ ܓܘ ܗܢܝ ܬܠܬܐ ܐܬ̈ܘܬܐ، ܘܩܪܘܝܐ ܟܓܒܐ ܠܦܘܬ ܣܝܩܐ ܕܠܐ ܚܘܪܐ. ܚܘܫܒܢܐ ܠܐ ܡܨܐ. ܓܘ ܡܫܘܚܬܐ ܥܠ ܫܬܐ ܡܠܝܘܢ̈ܐ ܕܡ̈ܠܐ ܕܡܘܕܥܢܘܬܐ (press)، ܡܢܝܢܐ ܕܦܘܫ̈ܩܐ ܡܘܪܦܘܠܘܓ̈ܝܐ ܕܡܨܝ ܗܘܘ ܩܐ token ܥܒܪܝܐ ܝܠܗ ܒܡܨܥܝܬܐ (mean) 2.4، ܩܘܒܠ 1.4 ܓܘ ܐܢܓܠܝܙܝܐ (Adler & Elhadad, 2006) – ܘܐܗܐ ܝܠܗ̇ ܥܠܬܐ ܕܩܐ ܟܠ ܨܢܘܪܐ ܥܒܪܝܐ ܐܝܬ ܕܪܓܐ ܕܠܝܬ ܓܘ ܐܢܓܠܝܙܝܐ: ܦܘܪܩܐ ܕܬܪܝܢܘܬܐ ܕܣܘܟܠܐ (disambiguation)، ܕܓܘܗ ܡܘܕܝܠ (model) ܟܚܐܪ ܐܝܢܐ ܡܢ ܩܪ̈ܝܢܐ ܫܪܝܪܐ ܝܠܗ. ܟܡܐ ܛܒܐ ܝܠܗ ܚܐܪܬܐ ܐܕܝܘܡ؟ ܡܘܕܝܠ ܥܒܪܝܐ ܛܒܐ ܡܢ ܟܠ ܕܡܘܕܥܐ ܝܠܗ، DictaBERT-Parse-large، ܥܠ ܒܘܚܪܢܐ ܕ-treebank ܟܡܛܐ ܠ-97.9% ܓܘ ܦܘܠܓܐ ܕܡ̈ܠܐ ܕܩܕܡ ܡܠܬܐ (prefixes)، 97.3% ܓܘ part-of-speech tagging ܘ-95.6% ܓܘ ܕܝܠܝ̈ܬܐ ܡܘܪܦܘܠܘܓ̈ܝܬܐ – ܐܠܐ ܒܣ ܠ-85.4% ܓܘ ܦܘܫܩܐ ܕ-syntax ܥܡ ܪ̈ܘܫܡܐ (LAS)، ܘܩܕܡ ܡܦܬܝܢܘܬܐ ܕ-corpus ܕܝܘܠܦܢܐ ܠ-975 ܐܠܦ̈ܐ ܕܡ̈ܠܐ ܡܢܝ̈ܢܐ ܩܠܝܠ ܢܚܝ̈ܬܐ ܗܘܘ (Shmidman, Shmidman, Koppel & Tsarfaty, 2024, Table 1). ܗܢܘ: ܒܕܪܓܐ ܕܡܠܬܐ ܠܥܠ ܡܢ 95%، ܒܕܪܓܐ ܕܦܬܓܡܐ ܒܨܝܪ؛ ܘܥܠ ܡ̈ܠܐ ܕ-10,000 ܡ̈ܠܐ ܐܦ 2-5% ܡܐܐ̈ ܕܛܥܝ̈ܘܬܐ ܝܢܐ ܕܟܥܐܠܝ ܠܡܢܝܢܐ.

ܒ. ܡܘܪܦܘܠܘܓܝܐ ܥܬܝܪܬܐ ܝܠܗ̇، ܘܡܠܬܐ "ܟܡܬܚܠܛܐ". ܡ̈ܠܐ ܕܣܡܝܟܘܬܐ (prepositions)، ܕܐܣܘܪܐ ܘܕܝܕܥܬܐ (ו، ה، ב، כ، ל، מ، ש) ܟܡܕܒܩܝ ܩܕܡ ܡܠܬܐ، ܘܫܡ̈ܗܐ ܕܩܢܝܢܐ ܘܕܡܐ ܕܥܠܗ ܥܒܕܐ (object) – ܒܬܪܗ̇: בביתם = ב + בית + הם؛ ראיתיו = ראיתי + אותו. ܓܘ treebank ܥܒܪܝܐ (ܠܘܚܐ 3)، 114,648 tokens ܐܝܬ ܓܘܝܗܝ 160,195 ܡ̈ܠܐ ܕ-syntax – ܗܢܘ ܚܕ ܡܢ ܟܠ ܬܠܬܐ tokens ܓܘ ܡ̈ܠܐ ܕܡܘܕܥܢܘܬܐ ܥܡܝ̈ܐ ܣܢܝܩ ܝܠܗ ܕܡܬܦܠܓ ܩܕܡ ܕܡܨܝܚ ܕܡܢܚ ܠܗ. ܡܐܢܐ ܕܟܫܘܪ ܥܠ ܦܘܠܓܐ ܒܕ ܡܢܐ "בבית"، "לבית"، "הבית" ܘ"בית" ܐܝܟ ܐܪܒܥܐ ܡ̈ܠܐ – ܘܡܠܬܐ ܕܡܬܚܙܝܐ ܣܓܝ "ש" (שֶ) ܠܐ ܒܕ ܚܙܐ ܠܗ̇ ܡܢ ܟܠ ܥܠܡ، ܡܛܠ ܕܐܝܗܝ̇ ܡܢ ܟܠ ܥܠܡ ܠܐ ܟܩܝܡܐ ܒܚܘܕܝܗ̇.

ܓ. ܫܪܫܐ ܘܡܫܩܠܐ: ܡܘܪܦܘܠܘܓܝܐ ܕܠܐ ܫܫܠܬܐ (non-concatenative). ܓܘ ܐܢܓܠܝܙܝܐ ܡܠܬܐ ܟܡܬܒܢܝܐ ܒܫܫܠܬܐ – walk + ed: morphemes ܦܪ̈ܝܫܐ ܕܟܡܬܚܠܛܝ ܥܠ ܣܦܪܐ ܕܫܬܐܣܐ، ܘܡܛܠ ܗܕܐ، ܐܝܟ ܕܐܡܪܠܗ McCarthy، ܡܨܝܚ ܕܡܬܥܒܕܝ ܬܘܒ "ܒܦܘܪܩܐ ܡܢ ܣܡܠܐ ܠܝܡܝܢܐ ܕܡ̈ܠܐ ܒܒܨܝܐ ܕܣܕܪ̈ܐ ܡܢܬܢ̈ܝܐ ܕܟܕܥܪܝ" (McCarthy, 1981, p. 373). ܓܘ ܥܒܪܝܐ ܘܓܘ ܥܪܒܝܐ ܫܪܫܐ "ܩܝܢܬܐ" ܝܠܗ̇ ܕܐܬ̈ܘܬܐ ܠܐ ܡܬܩܠ̈ܐ (consonants) ܥܠ ܬܚܬܝܬܐ ܦܪܝܫܬܐ، ܕܟܡܬܬܣܝܡܐ ܓܘ ܛܘܦܣܐ prosodic ܕܕܘܟ̈ܝܬܐ ܕܐܬ̈ܘܬܐ ܘܕܢܘ̈ܩܙܐ (ܦܐ̈ܬܐ 386-387): כ-ת-ב "ܡܙܩܪܐ ܝܠܗ" ܓܘ ܡܫܩܠܐ – כָּתַב، כּוֹתֵב، מִכְתָּב، הִכְתִּיב، כְּתוּבָה، הִתְכַּתֵּב – ܘ-McCarthy ܟܦܬܚ ܡܐܡܪܐ ܕܝܠܗ ܚܬܝܬܐܝܬ ܒܛܘܦܣܐ (paradigm) ܐܝܟ ܐܗܐ ܓܘ ܥܪܒܝܐ (kataba، kattaba، kitaabun، maktabun…)، ܐܣܟܝ̈ܡܐ ܕ"ܐܣܝܪ̈ܐ ܝܢܐ ܚܕ ܠܚܕ ܒܡܘܪܦܘܠܘܓܝܐ ܐܦ ܐܢ ܠܐ ܡܫܘܬܦܝ ܣܕܪ̈ܐ ܦܪ̈ܝܫܐ ܕܩܠ̈ܐ" (ܦܐ̈ܬܐ 374-375). ܡܢ ܐܗܐ ܠܝܬ ܣܕܪܐ ܡܫܘܬܦܐ ܕܐܬ̈ܘܬܐ ܕܡܨܝܚ ܕ"ܦܣܩܚ" (stemming)، ܘܡܐܢ̈ܐ ܕܟܫܩܠܝ ܫܘ̈ܠܡܐ ܕܡ̈ܠܐ – ܛܘܦܣܐ ܓܘ ܐܢܓܠܝܙܝܐ – ܕܠܐ ܣܘܟܠܐ ܝܢܐ ܓܘ ܥܒܪܝܐ. ܦܘܫܪܐ ܝܠܗ lemmatization (lemmatization) ܥܡ ܠܟܣܝܩܘܢ، ܘܐܦ ܐܝܗܝ̇ ܟܕܥܪܐ lemma (כתב، מכתב، כתובה – ܐܝܟ ܬܠܬܐ lemmas)، ܠܐ ܫܪܫܐ. ܘܓܘ ܡܡܠܠ̈ܢܐ ܫܪܫܐ ܬܚܬܝܬܐ ܫܪܝܪܬܐ ܝܠܗ̇، ܠܐ ܒܣ ܬܚܘܝܬܐ ܕܓܪܡܛܝܩܐ: Ravid ܘ-Schiff ܟܟܢܫܝ ܕܫܪܫܐ "ܟܐܚܕ ܕܪܓܐ ܦܪܝܫܐ ܕܚܘܝܐ ܓܘ ܠܟܣܝܩܘܢ ܕܗܘܢܐ" ܕܡܡܠܠ̈ܢܐ ܕܥܒܪܝܐ ܘܕܥܪܒܝܐ، ܕܟܛܥܢ ܣܘܟܠܐ ܕܠܒܐ ܕܟܐܣܪ ܟܠ ܒ̈ܢܝ ܫܪܒܬܐ ܡܘܪܦܘܠܘܓܝܬܐ (ג-ד-ל: גדל، גידל، הגדיל، גידול، מגדל، מגדלת، גדול…)، ܘܕܝܠܘ̈ܕܐ ܟܡܦܩܝ ܫܪ̈ܫܐ ܡܢ ܡ̈ܠܐ ܡܢ ܩܕܡ ܒܥܘܡܪܐ ܕܬܪܬܝܢ ܫܢ̈ܐ (Ravid & Schiff, 2006, pp. 790-791)؛ ܓܘ ܢܣܝ̈ܘܢܐ ܕܝܠܗܘܢ ܥܠ 300 ܝܠܘ̈ܕܐ ܕܡܕܪܫܬܐ، ܡܫܬܘܕܥܢܘܬܐ ܕܫܪܫܐ ܗܘܝܐ ܗܘܐ "ܪܫܝܐ ܘܩܝܡܐ" ܡܢ ܬܪܘܝܗܝ، ܘܣܘܟܠܐ ܕܡܫܩܠܐ ܬܒܝܪܐ ܝܠܗ ܝܬܝܪ ܘܟܡܬܚܙܐ ܒܬܪ – ܡܢ ܒܝܢܝ ܥܠ̈ܠܬܐ ܡܛܠ ܕܐܝܬ ܩܪܝܒ 4,000 ܫܪ̈ܫܐ ܩܘܒܠ ܩܪܝܒ 200 ܡܫܩ̈ܠܐ ܒܝܫ ܡܢ ܟܠ، ܘܡܛܠ ܕܣܘܟܠܐ ܕܡܫܩܠܐ ܠܐ ܝܠܗ ܫܘܝܐ (ܦܐ̈ܬܐ 808-809). ܫܘܐܠܐ ܕܒܘܨܝܐ ܥܠ "ܚܩܠܐ ܕܣܘܟܠܐ ܕܫܪܫܐ" ܣܡܝܟܐ ܝܠܗ ܗܕܟܐ ܥܠ ܚܕܝܘܬܐ ܕܢܦܫܐ ܫܪܝܪܬܐ – ܐܠܐ ܟܒܥܐ ܬܚܬܝܬܐ ܝܬܝܪܬܐ ܠܥܠ ܡܢ lemma، ܕܠܝܬ ܠܗ̇ ܡܐܢܐ ܡܛܝܒܐ.

ܕ. ܐܪܒܥܐ ܕܪ̈ܓܐ ܕܠܫܢܐ، ܐܪܒܥܐ ܠܫܢ̈ܐ. ܦܘܠܓܐ ܡܩܒܠܐ ܓܘ ܝܕܥܬܐ ܕܠܫܢܐ ܥܒܪܝܐ ܝܠܗ ܠܐܪܒܥܐ ܙܒ̈ܢܐ – ܕܟܬܒܐ ܩܕܝܫܐ، ܕܪ̈ܒܢܐ (Hazal)، ܕܕܪ̈ܐ ܡܨܥ̈ܝܐ ܘܐܝܣܪܐܝܠܝܐ – ܥܡ ܦܘܠܓ̈ܐ ܙܥܘܪ̈ܐ ܘܕܪ̈ܓܐ ܕܥܒܪܬܐ: ܕܟܬܒܐ ܩܕܝܫܐ ܠܥܬܝܩܬܐ، ܠܩܠܐܣܝܩܝܬܐ ܘܠܐܚܪܝܬܐ (ܕܒܬܪ ܓܠܘܬܐ)، ܘܕܪ̈ܒܢܐ ܠܠܫܢܐ ܕܬܢܐ̈ܝܐ (Tannaim) ܘܠܠܫܢܐ ܕܐܡܘܪ̈ܝܐ (Amoraim) (Pérez Fernández, 1997, p. 1). ܐܢܝ ܠܐ ܝܢܐ ܐܣܟܝ̈ܡܐ ܕܚܕ ܠܫܢܐ ܐܠܐ ܕܪ̈ܓܐ ܕܫܚܠܦܠܗܘܢ ܐܦ ܐܝܟ ܠܫܢܐ ܕܡܡܠܠܐ: ܠܫܢܐ ܕܟܬܒܐ ܩܕܝܫܐ، "ܕܒܫܬܐܣܐ ܠܫܢܐ ܕܣܦܪܘܬܐ ܗܘܐ، ܫܒܩܠܗ̇ ܕܗܘܝܐ ܠܫܢܐ ܚܝܐ ܓܘ ܙܒܢܐ ܕܒܬܪ ܓܠܘܬܐ ܕܒܒܠ"، ܘܠܫܢܐ ܕܚܟܝ̈ܡܐ ܗܘܐ ܡܡܠܠܐ ܕܟܠ ܝܘܡ ܓܘ ܝܗܘܕ ܗܠ ܚܢܩܐ ܕܡܪܘܕܘܬܐ ܕ-Bar Kokhba (135 ܡܫܝܚܝܐ) ܘܒܬܪ ܗܕܐ ܒܣ ܠܫܢܐ ܕܟܬܝܒܬܐ ܓܢܒ ܐܪܡܝܐ (Sáenz-Badillos, 1993, pp. 166, 171). ܦܘܪ̈ܫܢܐ Pérez Fernández ܟܡܢܐ ܠܗܘܢ ܓܘ ܫܘܪܝܐ ܕܟܬܒܐ ܕܝܠܗ ܐܝܟ ܡܐ ܕܟܡܬܗܪ ܟܠ ܩܪܘܝܐ ܕܡ̈ܠܐ ܕܪ̈ܒܢܐ: ש- ܒܕܘܟܬܐ ܕ-אשר، "של" ܒܕܘܟܬܐ ܕܣܡܝܟܘܬܐ (construct state)، "היה" ܥܡ ܒܝܢܝܐ (participle)، ܛܫܝܬܐ ܕ-vav ܕܗܘܦܟܐ، ܐܒܕܢܐ ܕ-infinitive absolute ܘܕܐܣܟܝ̈ܡܐ ܦܪ̈ܝܫܐ ܕܦܘܩܕܢܐ (imperative)، ܘ-nun ܒܕܘܟܬܐ ܕ-mem ܓܘ ܫܘܠܡܐ ܕܣܓܝܐܐ (ܦܐܬܐ 1)؛ ܘܒܣ ܩܪܝܒ ܦܠܓܐ ܡܢ ܡ̈ܠܐ ܕܠܫܢܐ ܕܬܢܐ̈ܝܐ ܡܫܘܬܦܐ ܝܠܗ ܥܡ ܟܬܒܐ ܩܕܝܫܐ – ܫܪܟܐ ܐܪܡܝܐ، ܝܘܢܝܐ، ܘܒܡܫܘܚܬܐ ܒܨܝܪܬܐ ܠܛܝܢܝܐ، ܐܟܕܝܐ ܘܦܪܣܝܐ (ܦܐܬܐ 4). ܥܒܪܝܐ ܕܕܪ̈ܐ ܡܨܥ̈ܝܐ ܠܫܢܐ ܕܟܬܝܒܬܐ ܗܘܐ ܘܠܐ ܕܡܡܠܠܐ، ܕܡܚܝܐ ܗܘܐ ܡܢ ܕܪܐ ܥܣܝܪܝܐ ܒܡܥܒܕܢܘܬܐ ܕܓܪܡܛܝܩܐ ܥܪܒܝܬܐ ܘܦܬܝܐ ܗܘܐ ܠܙܡܝܪܬܐ ܥܠܡܢܝܬܐ، ܠܦܝܠܘܣܘܦܘܬܐ ܘܠܝܕܥܬܐ (Sáenz-Badillos, pp. 202-203)؛ ܘܥܒܪܝܐ ܐܝܣܪܐܝܠܝܐ ܝܠܗ̇، ܠܦܘܬ Rosén (1958, as cited in Sáenz-Badillos, pp. 275-276)، koine ܕܓܘܗ̇ ܫܬܐܣ̈ܐ ܕܟܬܒܐ ܩܕܝܫܐ ܘܕܪ̈ܒܢܐ ܟܡܦܠܚܝ ܚܕ ܓܢܒ ܚܕ ܐܝܟ ܬܪܝܢ ܕܪ̈ܓܐ ܕܐܣܟܝܡܐ، ܥܠ ܫܬܐܣܐ ܠܐ ܥܒܪܝܐ ܘܥܡ syntax ܘܡ̈ܠܐ ܕܩܒܠܠܗܘܢ ܡܥܒܕܢܘܬܐ ܡܢ ܣܠܐܒܝܐ، ܡܢ ܓܪܡܢܝܐ ܘܝܬܝܪ ܡܢ ܐܢܓܠܝܙܝܐ (ܦܐܬܐ 280) – "ܡܐ ܕܓܪܡܛܝܩܐ ܡܫܠܡܢܝܬܐ ܚܙܝܐ ܗܘܐ ܐܝܟ ܛܥܝ̈ܘܬܐ ܝܢܐ، ܒܫܪܪܐ، ܚܘܕ̈ܬܐ ܕܕܪܓܐ ܚܕܬܐ ܓܘ ܠܫܢܐ" (ܦܐܬܐ 276). ܡܥܒܕܢܘܬܐ ܩܐ ܚܘܫܒܢܐ: ܡܘܕܝܠ (model) ܕܡܠܦܐ ܝܠܗ ܥܠ Haaretz ܘ-Wikipedia (AlephBERT، DictaBERT) ܠܐ ܒܕ ܡܨܐ ܥܠ ܦܐܬܐ ܕ-Gemara؛ ܩܐ ܐܗܐ ܒܢܝܐ ܝܠܗ BEREL (ܠܘܚܐ 2)؛ ܘܩܐ ܟܬܒܐ ܩܕܝܫܐ ܐܝܬ ܓܙܐ ܕܝܗܝ̈ܒܬܐ ܪܫܝܡܐ ܒܐܝܕܐ ܕܝܠܗ (BHSA, Table 4) ܕܠܐ ܣܢܝܩ ܝܠܗ ܠܡܘܕܝܠ ܡܢ ܟܠ ܥܠܡ. ܓܘܒܝܐ ܩܕܡܝܐ ܕܒܨܘܝܐ ܠܐ ܝܠܗ "ܐܝܢܐ ܡܐܢܐ" ܐܠܐ "ܐܝܢܐ ܥܒܪܝܐ".

ܒܘܨ̈ܝܐ ܫܪ̈ܫܝܐ ܕܚܩܠܐ: Wintner (2004) – ܥܣܪܐ ܫܢ̈ܐ ܩܕܡ̈ܝܐ ܕܝܕܥܬܐ ܕܠܫܢܐ ܚܘܫܒܢܝܬܐ ܥܒܪܝܬܐ؛ Itai & Wintner (2008) – ܡܒܘ̈ܥܐ ܕܒܢܝ̈ܐ ܗܘܘ ܓܘ Technion ܘܕܥܠܝܗܝ ܩܝܡ ܟܠ ܡܐ ܕܐܬܐ ܒܬܪ؛ ܘܡܐ ܕܫܚܠܦܠܗ ܓܘ ܥܣܪܐ ܫܢ̈ܐ ܕܢܝܘܪܘܢ̈ܐ (neural) – Tsarfaty, Bareket, Klein & Seker (2020).

2. ܐܘܠܨܢܐ ܕܠܐ ܟܡܬܚܙܐ: ܪܡܙܢܘܬܐ (encoding)، ܢܘ̈ܩܙܐ ܘܐܘܪܚܐ ܕܟܬܝܒܬܐ

ܩܕܡ ܕܡܡܠܠܚ ܥܠ ܡܘܪܦܘܠܘܓܝܐ، ܐܝܬ ܬܚܬܝܬܐ ܕܓܘܗ̇ ܥܒܪܝܐ ܟܬܒܪܐ ܕܠܐ ܕܚܕ ܐܢܫܐ ܚܐܪ – ܘܪܘܒܐ ܕ"gibberish" ܕܒܨ̈ܘܝܐ ܟܦܓܥܝ ܒܗ ܟܐܬܐ ܡܢܗ̇.

ܪܡܙܢܘܬܐ (encoding). ܦܐܝܠ ܥܒܪܝܐ ܝܠܗ UTF-8 ܝܢ windows-1255 (ܪܡܙܢܘܬܐ ܥܬܝܩܬܐ ܕ-Windows ܒܥܒܪܝܐ، ܕܬܘܒ ܟܦܠܛܐ ܡܢ SPSS ܘܡܢ ܬܘܟ̈ܢܝܬܐ ܥܬܝ̈ܩܬܐ). ܦܐܝܠ 1255 ܕܦܬܝܚܐ ܝܠܗ ܐܝܟ UTF-8 ܟܡܬܚܙܐ ܐܝܟ ܣܕܪܐ ܕܐܬ̈ܘܬܐ ܕܠܐ ܣܘܟܠܐ؛ ܦܐܝܠ UTF-8 ܕܠܐ BOM ܕܦܬܝܚܐ ܝܠܗ ܒܝܕ ܟܬܒܘܬܐ ܕܪܡܙܐ (library) ܕܟܣܒܪܐ ܪܡܙܢܘܬܐ ܠܛܝܢܝܬܐ – ܗܘ ܡܕܡ. ܦܐ̈ܬܐ ܕܡܚܫܒ̈ܢܐ ܕܣܐܝܛ ܟܬܪܨܝ ܬܪܬܝܢ ܐܝܟܢܝ̈ܘܬܐ ܒܚܘܕܝܗܝ (ܚܙܝܡܘܢ ܨܘܪܬܐ ܕܡܐܢ̈ܐ، ܦܣܩܐ 7)؛ ܡܐܢܐ ܕܠܒܪ – ܠܐ ܡܢ ܟܠ ܐܡܢ. ܢܡܘܣܐ: ܢܛܪܬܐ ܕܟܠ ܡܕܡ ܐܝܟ UTF-8، ܘܒܘܚܪܢܐ ܕܣܕܪܐ ܩܕܡܝܐ ܒܥܝܢܐ ܩܕܡ ܟܠ ܦܘܫܩܐ. ܐܘܠܨܢܐ ܕܠܐ ܣܓܝ ܡܬܚܙܐ: "ܥܒܪܝܐ ܚܙܝܬܐ" (visual Hebrew؛ ISO-8859-8 ܥܬܝܩܐ)، ܕܓܘܗ̇ ܐܬ̈ܘܬܐ ܢܛܝܪ̈ܐ ܝܢܐ ܒܛܟܣܐ ܕܚܘܝܐ، ܡܢ ܝܡܝܢܐ ܠܣܡܠܐ – ܡ̈ܠܐ ܟܡܬܚܙܝ ܗܦܝ̈ܟܐ ܒܬܪ ܫܘܢܝܐ.

ܢܘ̈ܩܙܐ ܘܐܬ̈ܘܬܐ ܕܩܝܢܬܐ (cantillation). ܓܘ Unicode، ܐܬܐ ܕܢܘܩܙܐ ܐܬܐ ܦܪܝܫܬܐ ܝܠܗ̇ ܕܟܡܬܚܠܛܐ ܥܡ ܐܬܘܬܐ (combining mark): ܓܘ ܒܠܘܩ ܥܒܪܝܐ (U+0590 ܗܠ U+05FF) ܐܝܬ 51 ܐܬ̈ܘܬܐ ܐܝܟ ܐܗܐ – ܛܥܡ̈ܐ (te'amim)، ܢܘ̈ܩܙܐ، dagesh ܘܢܘ̈ܩܙܐ ܕ-shin ܘ-sin – ܡܢ U+0591 (etnahta) ܗܠ U+05C7 (qamats qatan)، ܓܢܒ 27 ܐܬ̈ܘܬܐ (U+05D0 ܗܠ U+05EA) ܘܐܬ̈ܘܬܐ ܕܢܘܩܙܬܐ (Unicode Consortium, 2026، ܒܠܘܩ Hebrew). "שָׁלוֹם" ܥܡ ܢܘ̈ܩܙܐ ܘ"שלום" ܕܠܐ ܢܘ̈ܩܙܐ ܬܪܝܢ ܣܕܪ̈ܐ ܦܪ̈ܝܫܐ ܝܢܐ، ܘܠܣܕܪܐ ܥܡ ܢܘ̈ܩܙܐ ܐܝܬ ܝܬܝܪ ܡܢ ܚܕ ܐܣܟܝܡܐ ܕܢܛܪܬܐ: ܛܟܣܐ ܕܐܬ̈ܘܬܐ ܥܠ ܗܝ̇ ܐܬܘܬܐ ("ש + qamats + shin dot" ܩܘܒܠ "ש + shin dot + qamats") ܘܐܬ̈ܘܬܐ ܕܚܘܝܐ ܡܪܟ̈ܒܬܐ (presentation forms) ܕܒܠܘܩ U+FB1D ܗܠ U+FB4F (46 ܐܬ̈ܘܬܐ، ܡܬܠܐ שׁ ܐܝܟ ܚܕܐ ܐܬܘܬܐ). Normalization (NFC ܝܢ NFD) ܟܦܫܪ ܬܪܘܝܗܝ – ܐܗܘ ܟܦܪܩ ܐܬ̈ܘܬܐ ܕܚܘܝܐ ܘܟܛܟܣ ܐܬ̈ܘܬܐ ܠܦܘܬ ܓܢܣܐ ܕܚܘܠܛܐ ܕܝܠܗܝ (combining class) (Whistler, 2026، ܐܠܓܘܪܝܬܡ ܕܛܟܣܐ ܩܢܘܢܝܐ). ܩܕܡ ܡܢܝܢܐ – normalization، ܘܒܬܪ ܗܕܐ ܫܩܠܬܐ ܕܢܘ̈ܩܙܐ ܠܦܘܬ ܓܢܣܐ (Mn، ܐܬܐ ܕܠܐ ܦܬܝܘܬܐ) ܘܠܐ ܠܦܘܬ ܦܬܝܘܬܐ ܕܡܢܝ̈ܢܐ: ܦܬܝܘܬܐ U+0591 ܗܠ U+05C7 ܐܝܬ ܓܘܗ̇ ܐܦ maqaf (U+05BE)، paseq، sof pasuq ܘ-nun hafukha – ܘܐܝܢܐ ܕܟܥܒܪ ܦܬܝܘܬܐ ܟܥܒܪ ܐܦ ܐܢܝ. ܡ̈ܠܐ ܕܟܬܒܐ ܩܕܝܫܐ ܥܡ ܢܘ̈ܩܙܐ ܘܛܥܡ̈ܐ ܕܠܐ normalization ܒܕ ܝܗܒܝ ܩܐ ܟܠ ܡܠܬܐ ܣܓܝܐܘܬܐ (frequency) 1.

ܟܬܝܒܬܐ ܡܠܝܬܐ ܘܚܣܝܪܬܐ، ܐܬ̈ܘܬܐ ܕܫܘܠܡܐ، ܢܘܩܙܬܐ ܥܒܪܝܬܐ. "תוכנית" ܘ"תכנית"، "אמיתי" ܘ"אמתי"، "ליברלי" ܘ"ליבראלי" – ܗܝ̇ ܡܠܬܐ، ܬܪܝܢ ܣܕܪ̈ܐ، ܬܪܬܝܢ ܣܓܝܐܘ̈ܬܐ. ܢܡܘ̈ܣܐ ܕܟܬܝܒܬܐ ܡܠܝܬܐ ܕܐܩܕܡܝܐ ܕܠܫܢܐ ܥܒܪܝܐ (2017) ܠܐ ܫܠܡܠܗܘܢ ܦܘܠܓܐ – ܒܣ ܫܚܠܦܠܗܘܢ ܐܘܪܚܐ ܕܝܠܗ. geresh ܘ-gershayim ܥܒܪ̈ܝܐ (׳ ״) ܩܘܒܠ geresh ܘܢܘ̈ܩܙܐ ܕܡܐܡܪܐ ܕ-ASCII (' ") – "ד״ר" ܘ"ד"ר" ܬܪܝܢ tokens ܦܪ̈ܝܫܐ ܝܢܐ، ܘܗܕܟܐ ܐܦ ܟܠ ܩܘ̈ܨܪܐ (acronyms). maqaf ܥܒܪܝܐ (־) ܩܘܒܠ ܣܪܛܐ ܕ-ASCII (-). ܠܝܬ ܦܘܫܪܐ ܒܚܘܕܝܗ ܓܡܝܪܐ؛ ܐܝܬ normalization ܓܠܝܬܐ ܕܟܟܬܒܚ ܠܗ̇.

ܐܘܪܚܐ ܕܟܬܝܒܬܐ. ܟܬܝܒܬܐ ܥܒܪܝܬܐ RTL ܝܠܗ، ܐܢܐ ܡܢܝܢ̈ܐ، ܫܡ̈ܗܐ ܕܡܫܬܚܠܦ̈ܢܐ (variables) ܒܐܢܓܠܝܙܝܐ ܘܐܬ̈ܘܬܐ ܕܦܘܣܩܐ ܒܚܪܬܗ ܠܐ ܝܢܐ: ܓܘ ܐܠܓܘܪܝܬܡ ܕܐܘܪܚܐ ܕܟܬܝܒܬܐ ܕ-Unicode ܐܬ̈ܘܬܐ ܥܒܪ̈ܝܬܐ ܡܢ ܙܢܐ R ܝܢܐ، ܡܢܝܢ̈ܐ ܡܢ ܙܢܐ EN، ܘܐܬ̈ܘܬܐ ܕܦܘܣܩܐ ܘܩܘܠܦ̈ܐ "ܡܨܥܝ̈ܐ" ܝܢܐ – ܟܫܩܠܝ ܐܘܪܚܐ ܚܝܠܬܢܐ ܕܟܬܝܒܬܐ ܕܬܪܝܢ ܓܢܒ̈ܐ ܕܝܗܘܢ، ܘܐܢ ܓܢܒ̈ܐ ܦܪ̈ܝܫܐ ܝܢܐ، ܐܘܪܚܐ ܕܦܣܩܐ (Goregaokar & Leroy, 2026, Sections 3.2, 3.3.5) – ܘܦܐܪܐ ܝܕܝܥܐ ܝܠܗ: ܩܘܠܦ̈ܐ ܗܦܝ̈ܟܐ، ܐܬܐ ܕܒܨܘܪܐ ܡܢ ܝܡܝܢܐ ܕܡܢܝܢܐ، ܫܡ̈ܗܐ ܕܡܫܬܚܠܦ̈ܢܐ ܕܟܬܒܪܝ ܓܘ ܨܘܪܬܐ. ܫܪܝܐ ܕܬܩܢܐ ܓܢܗ ܟܡܩܪܒ ܝܠܗ ܦܪܫܘܬܐ (isolate): "ܐܬ̈ܘܬܐ ܓܘ ܦܪܫܘܬܐ ܠܐ ܡܨܝܐ ܕܡܥܒܕܝ ܥܠ ܛܟܣܐ ܕܐܬ̈ܘܬܐ ܠܒܪ ܡܢܗ، ܘܗܕܟ ܐܦ ܠܦܘܬ ܗܦܟܐ"، ܘܦܪܫܘܬܐ ܟܠܗ̇ ܠܘܬ ܐܬܪܐ ܕܚܕܪܗ̇ ܐܝܟ ܐܬܐ ܡܨܥܝܬܐ ܟܥܒܕܐ (ܬܡܢ، ܦܣܩܐ 2.4). ܓܘ ܐܗܐ ܣܐܝܛ ܐܗܐ span.en ܝܠܗ (unicode-bidi: isolate) ܩܐ ܟܠ ܛܝܡܐ ܢܘܟܪܝܐ؛ ܓܘ R – ܡܫܩܠܢܐ ܓܪܦܝܐ AGG (ܚܙܝܡܘܢ ܦܐܬܐ ܕܨܘܪܬܐ ܕܒܕܪܐ، ܦܣܩܐ "ܘܕܟܝ ܟܥܒܕܚ ܐܗܐ ܒ-R"). ܓܘ ܡܐܢ̈ܐ ܐܝܟ AntConc ܝܢ Voyant ܥܒܪܝܐ ܟܦܠܚܐ – ܐܢܐ ܡܚܘܝܢܘܬܐ ܡܨܝܐ ܕܗܘܝܐ ܗܦܝܟܬܐ ܝܢ ܬܒܝܪܬܐ، ܘܠܐ ܦܘܫܩܐ (analysis). ܠܦܘܪܫܢܐ: ܡܚܘܝܢܘܬܐ ܬܒܝܪܬܐ ܠܝܠܗ̇ ܣܗܕܘܬܐ ܕܡܢܝܢܐ ܛܥܝܐ، ܘܡܢܝܢܐ ܛܥܝܐ (ܣܒܒ segmentation) ܠܐ ܟܡܬܚܙܐ ܓܘ ܡܚܘܝܢܘܬܐ.

3. ܨܢܘܪܐ (pipeline): ܡܢ segmentation ܠܦܘܫܩܐ ܕ-syntax – ܘܡܐܢ̈ܐ

ܨܢܘܪܐ ܬܪܝܨܐ ܕܦܘܠܚܢܐ ܕܥܒܪܝܐ ܟܥܒܕ، ܒܐܗܐ ܛܟܣܐ: segmentation (ܦܪܫܘܬܐ ܕ-prefixes ܘ-suffixes)، ܦܘܫܩܐ ܕ-morphology (ܟܠ ܩܪ̈ܝܢܐ ܕܡܨܝܐ ܕܗܘܝܐ ܩܐ ܟܠ segment)، disambiguation (ܓܘܒܝܐ ܕܩܪܝܢܐ ܬܪܝܨܐ ܓܘ ܣܘܟܠܐ ܕܡܐܡܪܐ)، tagging ܕ-part-of-speech ܘ-lemma، ܦܘܫܩܐ ܕ-syntax (ܡܢܝܐ ܝܠܗ subject ܕܡܢܝܐ) ܘܦܘܪܫܐ ܕ-named entities (ܫܡ̈ܗܐ ܕܐܢܫ̈ܐ، ܕܐܬܪ̈ܘܬܐ، ܕܬܘܩ̈ܢܐ). ܣܘܟܠܐ ܪܫܝܐ ܕܥܣܪܐ ܫܢ̈ܐ ܐܚܪ̈ܝܐ ܝܠܗ ܕܐܢܝ ܕܪ̈ܓܐ ܠܐ ܝܢܐ ܚܐܪ̈ܐ ܡܢ ܚܕ̈ܕܐ: segmentation ܬܪܝܨܐ ܕ"הבית" ܬܠܐ ܒ-syntax، ܘ-syntax ܬܠܐ ܒ-segmentation – ܘܡܛܠ ܗܕܐ ܡܘܕܝ̈ܠܐ ܛܒ̈ܐ ܟܥܒܕܝ ܠܗܘܢ ܐܟܚܕܐ، ܘܠܐ ܚܕ ܒܬܪ ܚܕ (More, Seker, Basmova &; Tsarfaty, 2019; Tsarfaty, Sadde, Klein &; Seker, 2019). ܫܪܒܬܐ ܕܗܫܐ – DictaBERT-joint – ܟܥܒܕܐ ܟܠܗܘܢ ܐܫܬܐ ܓܘ ܚܕ ܡܘܕܝܠ (model).

ܩܐ ܒܘܨܝܐ ܟܢܫܝܐ ܣܘܟܠܐ ܥܒܕܢܝܐ ܝܠܗ: ܠܝܬ "tokenizer ܩܐ ܥܒܪܝܐ" ܕܡܨܝܚ ܡܪܗܛܚ ܠܗ ܘܒܬܪܗ ܡܫܡܠܚ ܒܡܐܢܐ ܐܢܓܠܝܙܝܐ. segmentation ܘ-lemmatization ܓܪܓ ܐܬܝܐ ܡܢ ܨܢܘܪܐ ܥܒܪܝܐ، ܘܡܦܩܢܐ (output) ܕܝܠܗ – ܟܬܝܒܬܐ ܕܥܒܝܕܐ ܝܠܗ̇ lemmatization ܘ-segmentation – ܝܠܗ ܡܥܠܢܐ ܩܐ ܟܠ ܫܪܟܐ ܕܡܐܢ̈ܐ (topic modeling، concordance، ܠܟܣܝܩܘ̈ܢܐ).

4. ܡܘܕܝ̈ܠܐ ܕܠܫܢܐ ܩܐ ܥܒܪܝܐ – ܘܡܐ ܟܡܟܠܦܐ ܠܢ tokenization

ܡܘܕܝ̈ܠܐ ܕܠܫܢܐ ܡܠܦ̈ܐ ܡܢ ܩܕܡ (pretrained language models) ܐܢܝ ܝܢܐ ܕܟܡܨܝܐ ܨܢܘܪܐ ܐܟܚܕܝܐ، ܦܘܪܫܐ ܕ-named entities ܘ-classification ܕܪ̈ܓܫܐ. ܩܐ ܥܒܪܝܐ ܐܝܬ ܐܕܝܘܡ ܫܪܒܬܐ ܙܥܘܪܬܐ ܐܢܐ ܪܨܝܢܬܐ، ܟܠܗ̇ ܓܘ Hugging Face؛ ܠܘܚܐ 2 ܟܡܛܟܣ ܠܗ̇ ܠܦܘܬ ܙܢܐ ܕܠܫܢܐ ܘܥܒܕܐ. ܦܘܪܫܢܐ ܡܘܚܡܐ ܝܠܗ ܒܝܢܝ ܡܘܕܝ̈ܠܐ ܕܪܡܙܢܘܬܐ (BERT ܘܡܢ̈ܘܬܐ ܕܝܠܗ – ܟܦܫܩܝ ܟܬܝܒܬܐ ܕܐܝܬ) ܘܡܘܕܝ̈ܠܐ ܡܘܠܕ̈ܢܐ (DictaLM، ܘܡܘܕܝ̈ܠܐ ܪ̈ܒܐ ܬܓܪ̈ܝܐ – ܟܒܪܝ ܟܬܝܒܬܐ). ܩܐ ܦܘܫܩܐ ܟܡܝܝܐ ܓܘ ܒܘܨܝܐ ܣܢܝܩܘܬܐ ܐܝܬ ܠܩܕܡܝ̈ܐ؛ ܬܪ̈ܝܢܐ ܡܘܚ̈ܡܐ ܝܢܐ ܩܐ ܟܢܘܫܝܐ ܘܩܐ tagging ܕܥܘܕܪܢܐ، ܘܓܘ ܐܗܐ ܣܐܝܛ – ܠܒܪ ܡܢ ܢܡܘܣܐ "ܕܠܐ AI ܓܘ ܙܒܢܐ ܕܪܗܛܐ". ܡܐ ܝܠܗ ܡܘܕܝܠ ܐܝܟ ܗܢܐ ܡܢ ܓܢܒܐ ܛܟܢܝܩܝܐ – neural network ܥܡܝܩܬܐ ܕܡܠܦܬܐ ܝܠܗ̇ ܡܢ ܩܕܡ ܘܟܡܬܩܢܚ ܠܗ̇ (fine-tuning) – ܘܐܝܡܢ ܐܗܐ ܢܝܛܘܪܩ ܦܐܫ ܠܒܘܨܝܐ ܟܢܫܝܐ ܘܐܝܡܢ ܠܐ: ܦܐܬܐ ܕܝܘܠܦܢܐ ܕܡܟܝܢܐ (machine learning)، ܦܣܩܐ 6.

ܟܠܦܬܐ ܕ-tokenization. ܡܘܕܝܠ ܕܠܫܢܐ ܠܐ ܟܚܙܐ ܡ̈ܠܐ ܐܠܐ "ܩܛܥ̈ܐ" (subword tokens) ܕܦܣܝ̈ܩܐ ܝܢܐ ܠܦܘܬ ܣܓܝܐܘܬܐ (frequency) ܓܘ corpus ܕܝܘܠܦܢܐ. ܓܘ ܡܘܕܝܠ ܕܠܫܢ̈ܐ ܣܓܝ̈ܐܐ ܕܡܠܦܐ ܝܠܗ ܒܝܬܪܘܬܐ ܥܠ ܐܢܓܠܝܙܝܐ، ܡܠܬܐ ܥܒܪܝܬܐ ܟܡܬܦܠܓܐ ܠܩܛܥ̈ܐ ܒܝܫ ܣܓܝ̈ܐܐ ܡܢ ܡܠܬܐ ܐܢܓܠܝܙܝܬܐ ܕܐܦ ܗܘ ܐܘܪܟܐ – ܘܡܛܠ ܗܕܐ ܟܬܝܒܬܐ ܥܒܪܝܬܐ ܟܡܟܠܦܐ ܒܝܫ ܣܓܝ tokens، ܟܡܬܦܫܩܐ ܒܚܬܝܬܘܬܐ ܒܨܝܪܬܐ، ܘܓܘ ܡܘܕܝ̈ܠܐ ܬܓܪ̈ܝܐ ܐܦ ܟܡܟܠܦܐ ܒܝܫ ܙܘܙ̈ܐ ܩܐ ܐܦ ܗܘ ܣܘܟܠܐ (Rust, Pfeiffer, Vulić, Ruder &; Gurevych, 2021; Ahia et al., 2023). ܐܗܐ ܝܠܗ ܥܠܬܐ ܕܓܙܐ ܕܡ̈ܠܐ ܪܒܐ ܡܢ 52K ܓܘ AlephBERT ܠ-128K ܓܘ AlephBERTGimmel، ܘܕܡܘܕܝ̈ܠܐ ܡܘܠܕ̈ܢܐ ܩܐ ܥܒܪܝܐ ܒܨܒܝܢܐ ܟܡܪܘܚܝ ܓܙܐ ܕܡ̈ܠܐ ܥܒܪܝܐ. ܩܐ ܒܘܨܝܬܐ: ܡܘܕܝܠ ܕܒܢܝܐ ܝܠܗ ܩܐ ܥܒܪܝܐ ܛܒܐ ܝܠܗ ܡܢ ܡܘܕܝܠ ܕܠܫܢ̈ܐ ܣܓܝ̈ܐܐ ܕܥܣܪܐ ܓܐܝ̈ܐ ܪܒܐ ܝܠܗ – ܠܐ ܡܛܠ ܪܒܘܬܐ ܐܠܐ ܡܛܠ tokenization.

ܦܬܝܚܐ ܝܢ ܚܪܝܟܐ، ܓܘ ܒܝܬܐ ܝܢ ܪܚܘܩܐ. ܟܬܒܐ ܕܝܘܠܦܢܐ ܩܐ ܝܕ̈ܥܬܐ ܕܒܪܢܫܘܬܐ ܕܝܓܝܛܠܝ̈ܬܐ (digital humanities) ܟܡܩܪܒ ܬܪܝܢ ܣܪ̈ܛܐ ܩܐ ܓܘܒܝܐ ܕܡܘܕܝܠ ܕܠܫܢܐ، ܘܛܒܐ ܝܠܗ ܕܡܦܠܚܚ ܠܗܘܢ ܥܠ ܠܘܚܐ 2: ܡܘܕܝܠ ܚܪܝܟܐ ܕܟܦܠܚ ܓܘ ܣܝܪܒܪ ܕܫܪܟܬܐ ܟܡܥܢܐ ܕ"ܝܗܝ̈ܒܬܐ ܕܝܘܟܘܢ ܟܡܬܬܚܕܝ ܘܟܡܬܦܠܚܝ ܓܘ ܣܝܪ̈ܒܪܐ ܕܡܪ̈ܘܬܐ"، ܘܓܘ ܫܪ̈ܛܐ ܕܡܦܠܚܢܘܬܐ ܕܡܐܢ̈ܐ ܣܓܝ̈ܐܐ ܡܦܠܚܢܐ ܟܫܠܡ ܕ-prompts ܘܡܦܩ̈ܢܐ ܦܝܫܝ ܡܘܦܠܚ̈ܐ ܩܐ ܝܘܠܦܢܐ ܝܬܝܪܐ؛ ܚܠܦܬܐ ܝܠܗ̇ ܡܘܕܝܠ ܦܬܝܚܐ ܕܟܡܚܬܬܚ ܘܟܡܪܗܛܚ ܓܘ ܒܝܬܐ (ܡܬܠܐ، ܓܘ Ollama)، ܘܩܐ ܫܘܐ̈ܠܐ ܥܠ ܟܢܫܐ ܕܟܬܒ̈ܐ ܬܚܘܡܐ – ܛܟܣܐ ܕܡܦܩܬܐ ܕܙܢܐ RAG، ܕܟܦܢܐ ܒܣ ܡܢ ܓܙܐ ܕܒܘܨܝܬܐ ܬܚܡܐ ܝܠܗ̇، ܘܡܛܠ ܗܕܐ ܚܬܝܬܐ ܝܠܗ ܒܝܫ ܘܦܐܫ ܠܝܗܝ̈ܒܬܐ ܪܓܝܫ̈ܬܐ ܝܢ ܢܛܝܪ̈ܬܐ (Drucker & Albrezzi, 2026, pp. 118-119, 121). ܟܠ ܡܘܕܝ̈ܠܐ ܓܘ ܠܘܚܐ 2 ܡܢ ܙܢܐ ܩܕܡܝܐ ܝܢܐ – ܦܬܝܚ̈ܐ ܘܟܦܠܚܝ ܥܠ ܡܚܫܒܢܐ ܕܝܘܟܘܢ؛ ܩܐ ܟܬܝܒ̈ܬܐ ܕܡܫܐܠܬ̈ܐ (interviews) ܐܗܐ ܠܝܠܗ ܝܘܬܪܢܐ ܕܦܫܝܩܘܬܐ ܐܠܐ ܫܪܛܐ (ܚܙܝܡܘܢ "ܐܝܟܐ ܝܠܗ̇ ܟܬܝܒܬܐ ܕܝܘܟܘܢ؟" ܓܘ ܦܣܩܐ 7). ܘܙܘܗܪܐ ܐܚܪܢܐ ܡܢ ܐܦ ܗܘ ܨܚܚܐ: hallucinations ܗܠܝܐ ܟܗܘܝܐ، ܘܡܕܡ ܕܠܝܬ ܓܘ ܛܟܣܐ ܕܝܘܠܦܢܐ ܕܡܘܕܝܠ ܓܘܢܝܐ ܟܡܬܚܙܐ ܐܝܟ ܕܠܝܬ – ܡܦܩܢܐ ܕܡܘܕܝܠ ܡܘܠܕܢܐ ܟܡܬܒܚܪ ܠܩܘܒܠ ܡܒܘܥܐ، ܘܠܐ ܟܡܬܐܡܪ ܐܝܟ ܡܒܘܥܐ (ܦܐܬܐ 119، 121).

5. Corpora، treebanks ܘܓܙ̈ܐ ܕܝܗܝ̈ܒܬܐ

treebank (treebank) ܝܠܗ corpus ܕܓܘܗ ܒܢܝ̈ ܐܢܫܐ ܪܫܡܠܗܘܢ segmentation، tagging ܘ-syntax ܬܪ̈ܝܨܐ – ܐܗܐ ܝܠܗ "ܡܠܦܢܐ" ܕܟܠ ܡܘܕܝܠ، ܘܐܦ ܡܫܘܚܬܐ ܕܥܠܗ̇ ܟܡܫܚܚ ܠܗ. ܩܐ ܥܒܪܝܐ ܐܝܬ ܬܪܝܢ ܓܠܝ̈ܐ ܓܘ Universal Dependencies: HTB ܥܬܝܩܐ (ܨܚܝܦܘܬܐ ܕܫܢ̈ܐ ܕ-90) ܘ-IAHLTwiki ܚܕܬܐ (Wikipedia ܕܗܫܐ). ܙܥܘܪܘܬܐ ܕܝܗܘܢ – ܡܩܪܒܘܬ 5,000-6,000 ܡܐܡܪ̈ܐ ܟܠ ܚܕ، ܒܦܘܚܡܐ ܥܡ ܥܣܪ̈ܐ ܕܐܠܦ̈ܐ ܒܐܢܓܠܝܙܝܐ – ܝܠܗ̇ ܟܪܝܘܬܐ ܕܒܢܝܢܐ ܕܟܠܗ ܚܩܠܐ (Tsarfaty et al., 2020). ܓܢܒܝܗܘܢ: corpus ܕܡܬܠܐ ܕܡ̈ܠܐ ܕܡܕܝܢܝܘܬܐ – ܟܬܝ̈ܒܬܐ ܕ-Knesset – ܘ-corpora ܪ̈ܒܐ ܕ-web.

6. ܥܒܪܝܐ ܩܕܡ ܥܒܪܝܐ ܚܕܬܐ: ܟܬܒܐ ܩܕܝܫܐ، ܚܟܝܡ̈ܐ ܕܬܠܡܘܕ (Hazal)، ܕܪ̈ܐ ܡܨܥܝ̈ܐ

ܩܐ ܒܘܨܝܬܐ ܕܟܬܒܐ ܩܕܝܫܐ ܘܩܐ ܒܘܨܝܐ ܕܣܦܪܝܘܬܐ ܕܚܟܝܡ̈ܐ ܕܬܠܡܘܕ ܗܘܝܐ ܦܪܝܫܐ ܝܠܗ ܡܢ ܥܩܪܐ – ܘܒܚܕ ܐܘܪܚܐ ܛܒܐ ܝܠܗ ܒܝܫ. ܟܬܝܒܬܐ ܕܟܬܒܐ ܩܕܝܫܐ ܟܠܗ̇ ܥܡ tagging ܝܠܗ̇، ܡܠܬܐ ܡܠܬܐ، ܒܪܡܐ ܕܠܝܬ ܩܐ ܚܕ corpus ܚܕܬܐ: ܓܙܐ ܕܝܗܝ̈ܒܬܐ ܕ-ETCBC (Eep Talstra Centre for Bible and Computer ܓܘ Vrije Universiteit Amsterdam؛ ܗܠ 2013 – ܓܘܕܐ ܕ-informatics WIVU، ܕܫܬܐܣܬܗ̇ ܗܘܐ ܓܘ 1977؛ Kingham & van Peursen, 2018، ܦܐܬܐ 1-2) ܟܪܫܡ ܩܐ ܟܠ ܡܠܬܐ ܟܬܝܒܬܐ، morphology، part of speech ܘ-lexeme، ܩܐ ܟܠ phrase ܘܟܠ clause ܬܚܘ̈ܡܐ ܘܥܒܕܐ ܕܝܗܘܢ، ܘܩܐ ܟܠ clause ܐܦ ܙܢܐ ܕܝܠܗ̇، ܬܚܘܡܐ ܕ-discourse ܕܝܠܗ̇ ܘܕܘܟܬܗ̇ ܓܘ ܛܟܣܐ ܕ-clauses (ܬܡܢ، ܦܐܬܐ 3). ܥܠ 426,555 ܡ̈ܠܐ ܕ-Codex Leningradensis (ܠܦܘܬ BHS) ܝܬܝܒ̈ܐ ܝܢܐ ܡܩܪܒܘܬ 946 ܐܠܦ̈ܐ ܕܡܕܡ̈ܐ ܕܠܫܢܐ ܘ-22.6 ܡܠܝܘܢ̈ܐ ܕܛܝ̈ܡܐ ܕ-features (Roorda, 2017, Table 18.2, p. 222)، ܕܦܬܝ̈ܚܐ ܝܢܐ ܩܐ ܫܘܐ̈ܠܐ ܕ-syntax ܥܠ ܟܠܗ̇ ܟܬܒܐ ܩܕܝܫܐ ܥܒܪܝܐ – ܓܘ ܒܪܐܘܙܪ ܒܝܕ SHEBANQ (ܡܢ 2014) ܘܒ-Python ܒܝܕ Text-Fabric (ܡܢ 2017؛ Roorda, 2017، ܦܐܬܐ 223-224؛ Roorda, 2018). ܛܝܡܐ: ܡܘܕܝܠ ܕܠܫܢܐ ܕܒܣܬܪ tagging ܬܐܘܪܝܐ ܝܚܝܕܝܬܐ ܝܠܗ̇، ܘܦܐܪ̈ܐ ܕܒܘܨܝܐ ܕܟܣܡܟܝ ܥܠܗ̇ ܟܝܪܬܝ ܦܣܩ̈ܐ ܕܝܠܗ̇. ܡܐ ܒܚܬܝܬܘܬܐ ܟܝܪܬܝܚ – ܠܬܚܬ، ܡܢ ܦܘܡܐ ܕܒܢܘ̈ܝܐ ܕܓܙܐ ܓܢܗܘܢ.

"ܚܙ̈ܝܬܐ، ܠܐ ܬܐܘܪܝܐ" – ܘܬܚܘ̈ܡܐ ܕܐܗܐ ܡܐܡܪܐ. Roorda ܟܡܚܘܐ ܠܓܘܒܝܐ ܫܬܐܣܝܐ ܗܕܟܐ: ܓܘܕܐ "ܠܐ ܨܒܐ ܗܘܐ ܕܡܨܝܪ ܠܡ̈ܥܝܪܢܘܬܐ ܓܘ ܡܘܕܝܠ ܕܠܫܢܐ ܕܕܪ̈ܓܐ، ܡܛܠ ܕܨܒܐ ܗܘܐ ܕܡܚܘܐ ܚܙ̈ܝܬܐ، ܠܐ ܬܐܘܪܝܐ"، ܘܡܕܡ ܕܗܘܐ ܝܠܗ "ܓܙܐ ܕܝܗܝ̈ܒܬܐ ܒܣܘܟܠܐ ܕܝܗܝ̈ܒܬܐ ܕܚܙ̈ܝܬܐ ܕܥܠܝܗܝܢ ܡܨܝܐ ܕܡܫܬܐܣܝ ܬܐܘܪܝ̈ܐ" (Roorda, 2017, pp. 218-219). ܬܪܝܢ ܢܡܘ̈ܣܐ ܕܥܒܕܐ ܟܡܚܘܝ ܐܗܐ (Kingham & van Peursen, 2018, pp. 3-4; Talstra & Sikkel, 2000, pp. 54-56, Figure 7): ܡܢ ܐܣܟܝܡܐ ܠܥܒܕܐ – ܟܪܫܡܚ ܩܕܡܐ ܕܝܠܢܝܘ̈ܬܐ ܕܐܦܐ ܘܒܬܪܗ ܒܣ ܪܘ̈ܫܡܐ ܕܥܒܕܐ، ܕܠܐ ܐܣܘܪܐ ܠܬܐܘܪܝܐ ܕܠܫܢܐ ܗܠ ܒܬܪ ܟܢܫܐ ܕܝܗܝ̈ܒܬܐ؛ ܘܡܢ ܬܚܬ ܠܥܠ – ܡܢ morphemes ܠ-lexemes، ܠ-phrases، ܠ-clauses، ܠܡܐܡܪ̈ܐ، ܠܟܬܝܒܬܐ، ܒܦܘܪܫܐ ܕܛܘ̈ܦܣܐ ܕܡܦܠܚܢܐ ܟܡܫܪܪ ܘܬܘܟܢܝܬܐ ܟܝܠܦܐ ("ܓܪܡܛܝܩܝ ܠܐ ܦܝܫܐ ܣܝܡܬܐ ܡܢ ܩܕܡ، ܐܠܐ ܦܝܫܐ ܠܒܝܫܬܐ ܐܣܟܝܡܐ ܓܘ ܦܘܫܩܐ ܕܝܗܝ̈ܒܬܐ" – Talstra & Sikkel، ܦܐܬܐ 46). ܐܢܐ ܒܢܘ̈ܝܐ ܕܓܙܐ ܓܢܗܘܢ ܟܡܬܚܡܝ: "ܠܐ ܐܣܘܪܐ ܠܚܕ ܬܐܘܪܝܐ ܓܢܗ ܙܢܐ ܕܬܐܘܪܝܐ ܝܠܗ – ܚܕܐ ܕܟܦܠܚܐ ܒܣ ܐܢ ܒܢܝܢܐ ܘܥܒܕܐ ܓܘ ܟܬܝܒܬܐ ܐܣܝܪ̈ܐ ܝܢܐ ܬܪܝܨܐܝܬ" (Kingham & van Peursen, p. 4)؛ ܫܬܐܣ̈ܐ ܕܬܐܘܪܝܐ ܟܕܥܪܝ ܠܓܪܡܛܝܩܝ ܕܟܬܝܒܬܐ structuralist ܕ-Wolfgang Schneider ܘ-Harald Weinrich (ibid., Note 10; Talstra & Sikkel, p. 62, Note 26)؛ ܘ"ܠܐ ܡܨܝܐ ܕܗܘܝܐ ܡܥܠܢܐ 'ܡܨܥܝܐ'، ܝܢܝ ܕܠܐ ܚܘܫܒܐ ܕܬܐܘܪܝܐ، ܕܝܗܝ̈ܒܬܐ ܕܠܫܢܐ" – ܒܢܝܢܐ ܕܓܙܐ ܓܢܗ ܒܘܨܝܐ ܝܠܗ، ܘܠܐ ܒܣ ܫܬܐܣܐ ܩܐ ܒܘܨܝܐ (Talstra & Sikkel, p. 53). ܐܪܒܥܐ ܦܣܩ̈ܐ ܕܟܠ ܦܐܪܐ ܟܝܪܬ:

1. ܬܚܘ̈ܡܐ ܕ-clause. segmentation ܟܗܘܝܐ ܒܬܪܝܢ ܕܪ̈ܓܐ: ܩܕܡܐ ܠܚܕܝܘ̈ܬܐ ܕܦܪܣܐ (distribution) – "ܐܛܘ̈ܡܐ"، ܣܕܪ̈ܐ ܕܡ̈ܠܐ ܕܠܐ ܡܨܝܐ ܕܦܝܫܝ ܦܠܝܓ̈ܐ ܒܝܫ ܘܠܐ ܟܡܬܦܣܩܝ – ܘܒܬܪܗ ܒܣ ܠܚܕܝܘ̈ܬܐ ܕܥܒܕܐ، ܕܡܨܝܐ ܕܗܘܝܐ ܡܪܟܒ̈ܐ ܡܢ ܚܕܟܡܐ ܐܛܘ̈ܡܐ ܘܡܛܠ ܗܕܐ ܡܨܝܐ ܕܐܚܕܝ ܡܥܠܬܐ ܘܦܬ̈ܚܐ (Talstra & Sikkel, 2000، ܦܐܬܐ 47-48، 51؛ Kingham & van Peursen, 2018، ܦܐܬܐ 5-6؛ ܡܬܠܐ ܕܝܗܘܢ: ܒܪܝܬܐ ܓ 11، ܕܓܘܗ clause ܕ-"אשר" ܘ-clause ܕ-"לבלתי" ܦܪܫܝ ܒܝܢܝ ܬܪܝܢ ܡܢ̈ܘܬܐ ܕ-clause ܕܟܡܚܕܪܐ). ܦܣܩܐ ܩܫܝܐ ܝܠܗ participle: ܐܡܢ ܕ-morphology ܬܪܬܝܢܝܬܐ ܝܠܗ̇ ܘ-phrases ܩܪ̈ܝܒܐ ܡܨܝܐ ܕܡܬܦܫܩܝ ܐܝܟ ܬܠܝ̈ܐ ܒܝܗ، participle ܟܡܬܚܫܒ predicate ܒܪܡܐ ܕ-clause – ܘܡܛܠ ܗܕܐ "ארץ אכלת יושביה הוא" (ܡܢܝܢ̈ܐ ܝܓ 32) ܟܡܬܪܡܙ ܐܝܟ relative clause ܦܪܝܫܬܐ ܘܠܐ ܐܝܟ ܚܕ noun phrase (Talstra & Sikkel, pp. 49-51, 52). ܐܝܢܐ ܕܟܡܢܝܐ clauses، ܟܡܫܚܐ ܐܘܪܟܐ ܕ-clause ܝܢ ܟܚܫܒܐ ܐܣܘܪ̈ܐ ܕ-clauses، ܟܡܢܝܐ ܐܗܐ ܦܣܩܐ.

2. ܙܢܐ ܕ-clause ܘܬܚܘܡܐ ܕ-discourse. ܟܠ clause ܟܫܩܠܐ ܙܢܐ ܠܦܘܬ ܐܣܟܝܡܐ ܕܦܥܠܐ ܪܫܝܐ ܘܛܟܣܐ ܕܡܢ̈ܘܬܐ (wayyiqtol ܕܠܐ subject ܓܠܝܐ، w-X-qatal، ܡܐܡܪܐ ܫܡܢܝܐ، clause ܕ-participle ܘܫܪܟܐ)، "ܓܢܣܐ" (ܕܦܥܠܐ، ܕܫܡܐ، ܕܠܐ predicate) ܘܬܚܘܡܐ ܕ-discourse – ܬܫܥܝܬܐ، ܡܡܠܠܐ ܝܢ ܡܥܠܝܐ (Kingham & van Peursen, 2018, p. 3). ܬܚܘܡܐ ܕ-discourse ܟܡܬܦܣܩ ܠܦܘܬ ܦܥܠܐ ܘܠܐ ܠܦܘܬ ܩܪܝܢܐ ܕܣܦܪܝܘܬܐ: clause ܕ-wayyiqtol ܟܡܬܪܫܡܐ ܬܫܥܝܬܐ (N)، clause ܕܟܦܬܚܐ ܒ-"אמר" ܝܢ "דבר" ܟܦܬܚܐ ܡܬܠܐ ܡܢ ܡ̈ܠܐ (Q)، ܦܘܩܕܢܐ ܝܢ yiqtol ܕܠܐ ܦܬܚܐ ܕܡܡܠܠܐ – ܡܡܠܠܐ (D)، ܘܐܢ ܠܝܬ ܪܡܙܐ ܕܓܪܡܛܝܩܝ ܐܗܝ clause ܟܝܪܬܐ ܪܘܫܡܐ ܕ-clause ܕܠܥܠ ܡܢܗ̇؛ ܗܕܟ "QN" ܟܪܫܡ wayyiqtol ܓܘ ܡܡܠܠܐ ܬܪܝܨܐ، ܘ-"ND" – yiqtol ܓܘ ܬܫܥܝܬܐ (Talstra & Sikkel, 2000, p. 62). ܐܗܐ ܝܠܗ text-syntax ܕ-Weinrich ܘ-Schneider، ܘܡܛܠ ܗܕܐ "ܬܫܥܝܬܐ" ܘ-"ܡܡܠܠܐ" ܠܟܐ ܓܢ̈ܣܐ (categories) ܝܢܐ ܕܛܟܣܐ ܕܦܥܠܐ، ܠܐ ܕ-genre. ܥܒ̈ܕܐ ܕܡܢ̈ܘܬܐ (subject، predicate، object، complement) ܕܪܓܐ ܕܥܒܕܐ ܝܢܐ، ܬܪܝܢܐ، ܘܐܢܝ ܝܢܐ ܕܟܡܨܝܐ ܡܬܠܐ ܦܘܚܡܐ ܛܟܣܢܝܐ ܕ-valency ܕܦ̈ܥܠܐ (ܬܡܢ، ܦܐܬܐ 63).

3. ܘܐܘ ܕܚܘܝܕܐ ܘܛܟܣܐ ܕܕܪ̈ܓܐ ܕ-clauses. ܬܚܘܡܐ ܕ-clause ܟܡܬܦܣܩ، ܒܝܢܝ ܡܕܡ̈ܐ ܐܚܪ̈ܢܐ، ܐܡܢ ܕܒܬܪܝܢ ܓܢܒ̈ܐ ܕܘܐܘ ܐܝܬ phrases ܡܢ ܙܢ̈ܐ ܦܪ̈ܝܫܐ – prepositional phrase، ܘܐܘ، noun phrase – ܘܡܬܠܐ ܕܡܟܬܒ̈ܢܐ ܝܠܗ ܒܪܝܬܐ ܟܒ 1 (Kingham & van Peursen, 2018, p. 5). ܐܢܐ ܓܙܐ ܠܐ ܟܩܐܡ ܓܘ ܬܚܘܡܐ: ܟܒܢܐ ܛܟܣܐ ܕܕܪ̈ܓܐ ܕ-clauses "ܝܡܐ" ܘ-"ܒܪܬܐ". ܡܢܬܐ ܕܒܨܝܪ ܕܢܣܝܘܢܐ ܝܠܗ̇ ܟܣܡܟܐ ܥܠ ܡ̈ܠܐ ܕܚܘܝܕܐ ܘܡ̈ܠܐ ܕܐܣܘܪܐ ܓ̈ܠܝܐ؛ ܠܒܪ ܡܢܗ̇ ܬܘܟܢܝܬܐ ܟܒܨܝܐ ܕܡܝܘܬܐ – ܩܘܪܒܐ ܕܐܣܟܝ̈ܡܐ ܕܦ̈ܥܠܐ، ܐܡܝܢܘܬܐ ܕ-person، number ܘ-gender، ܘܣܓܝܐܘܬܐ (frequency) ܕܐܣܘܪ̈ܐ ܕܡܫܟܚܐ ܝܢܐ ܒܝܢܝ ܬܪܝܢ ܙܢ̈ܐ ܕ-clauses (ܬܡܢ، ܦܐܬܐ 6). Talstra ܘ-Sikkel ܟܡܚܘܝ ܐܗܐ ܐܝܟ ܬܘܪܨܐ ܕܬܪܬܝܢ ܚܠܦ̈ܬܐ ܕܡܣܠܝܐ ܝܢܐ: segmentation ܕܣܪܛܐ ܕ-Richter ܓܘ BHt، ܕܠܐ ܟܚܫܒ ܠܚܘܝܕܐ ܒܝܢܝ main clauses ܒ-ܘܐܘ ܡܢ ܟܠ (Talstra & Sikkel, 2000, pp. 39-40)، ܘܦܘܫܩܐ "ܡܢ ܠܥܠ ܠܬܚܬ" ܕܣܝܠܗܝܡܪ، ܕܟܦܪܫ ܪ̈ܡܐ ܕܟܬܝܒܬܐ ܠܦܘܬ ܡܫܘ̈ܚܬܐ ܕܣܦܪܝܘܬܐ ܘܠܐ ܕܠܫܢܐ ܘܟܫܒܩ clauses ܕ-participle ܘ-infinitive ܐܝܟ clauses (ܬܡܢ، ܦܐܬܐ 40-42). ܘܡܘܚܡܐ: ܛܟܣܐ ܕܕܪ̈ܓܐ ܕ-clauses ܘܦܣܩܐ ܪܫܝܡ̈ܐ ܝܢܐ ܓܘ ܓܙܐ ܐܝܟ features ܕܢܣܝܘܢܐ (Kingham & van Peursen, p. 3) – ܐܝܢܐ ܕܟܣܡܟ ܥܠܝܗܝܢ ܓܪܓ ܐܡܪ ܐܗܐ.

4. ܟܬܝܒܬܐ، transliteration ܘܪܡܙܢܘܬܐ (coding). ܟܬܝܒܬܐ ܝܠܗ̇ Codex Leningradensis ܠܦܘܬ ܡܦܩܬܐ ܕ-BHS (Roorda, 2017, p. 218)؛ morphology "ܪܡܝܙܬܐ ܝܠܗ̇ ܘܠܐ ܥܡ tagging" – prefixes، lexeme، ܙܘܥ̈ܐ ܡܦܪ̈ܫܢܐ ܘ-suffixes ܓܘ transliteration ܓܘܝܐ ܕ-ETCBC، ܕܡܢܗ ܬܘܟܢܝܬܐ ܟܦܠܛܐ ܠܦܘܫܩܐ (Verheij, 1994, as cited in Kingham & van Peursen, 2018, p. 4)؛ ܘ-Roorda ܓܢܗ ܟܟܬܒ ܕܡܪܟܙܐ "ܛܒܝܥܐ ܝܠܗ ܓܘ ܐܘܪ̈ܚܬܐ ܕܝܠܗ" – transliteration ܕܝܠܗ ܘܐܘܪ̈ܚܬܐ ܕܪܡܙܢܘܬܐ ܕܝܠܗ، ܕܟܬܝܒ̈ܐ ܝܢܐ ܩܐ ܥܡܐ ܒܣ ܥܡ ܦܬܚܐ ܓܘ 2014 (Roorda, 2017, p. 226). ܓܙܐ ܟܡܬܚܕܬ ܒܢܘ̈ܣܚܐ (ETCBC4, 4b, 4c; ibid., p. 221)، ܘ-Text-Fabric ܠܐ ܟܠܝܘܡ ܐܦ ܗܘ ܝܠܗ ܐܝܟ ܢܘܣܚܐ ܕܣܝܪܒܪ (Kingham & van Peursen, Note 26).

ܡܐ ܟܟܬܒܚ ܓܘ ܨܚܚܐ ܕܐܘܪܚܐ (method section). ܕ-segmentation ܠ-clauses، ܙܢ̈ܐ ܕ-clauses، ܬܚܘ̈ܡܐ ܕ-discourse ܘܛܟܣܐ ܕܕܪ̈ܓܐ ܕ-clauses ܦܐܪ̈ܐ ܝܢܐ ܕܡܘܕܝܠ ܕ-ETCBC – ܡܢ ܬܚܬ ܠܥܠ، ܡܢ ܐܣܟܝܡܐ ܠܥܒܕܐ، ܥܠ ܫܬܐܣ̈ܐ ܕܓܪܡܛܝܩܝ ܕܟܬܝܒܬܐ ܕ-Weinrich ܘ-Schneider – ܘܠܐ ܫܪ̈ܪܐ ܓܠܡ̈ܐ ܕܟܬܝܒܬܐ؛ ܐܝܢܐ ܢܘܣܚܐ ܕܝܗܝ̈ܒܬܐ ܡܘܦܠܚܐ ܝܠܗ (ܘܡܢܝܢܐ ܕ-BHSA ܓܘ GitHub ܝܢ DOI ܓܘ DANS)؛ ܘܐܝܢܝ features ܕܢܣܝܘܢܐ ܥܠ̈ܠܐ ܝܢܐ ܓܘ ܦܘܫܩܐ. ܒܢܘ̈ܝܐ ܕܓܙܐ ܓܢܗܘܢ ܟܚܙܝ ܠܗ "ܡܐܢܐ ܕܒܘܨܝܐ ܕܓܪܡܛܝܩܝ ܕܟܬܝܒܬܐ" ܘܠܐ ܦܐܪܐ ܐܚܪܝܐ (Talstra & Sikkel, 2000, p. 35) – ܘܒܘܨܝܬܐ ܕܟܡܝܬܝܐ ܡܬܠܐ ܡܢܗ ܟܝܪܬܐ ܐܦ ܐܗܐ ܐܬܪܐ. ܢܘܣܚܐ ܐܚܪܝܐ ܕܐܗܐ ܦܣܩܐ، ܐܝܟ ܟܠ ܦܣܩ̈ܐ ܓܘ ܦܐܬܐ، ܠܝܠܗ ܕ-ETCBC ܐܠܐ ܕܣܐܝܛ؛ ܐܝܢܐ ܕܟܦܠܚܐ ܓܘ ܒܘܚܪܢܐ ܚܘܫܒܢܝܐ ܕܟܬܒܐ ܩܕܝܫܐ (computational biblical criticism) ܦܝܫܐ ܪܚܡܐ ܕܡܢܣܚܐ ܠܗ ܡܢ ܥܒܕܐ ܕܝܠܗ̇.

ܩܐ ܣܦܪܝܘܬܐ ܕܚܟܝܡ̈ܐ ܕܬܠܡܘܕ ܠܝܬ tagging ܓܡܝܪܐ ܒܐܝܕܐ، ܘܡܛܠ ܗܕܐ ܬܡܢ ܥܠܠܐ ܝܠܗ ܡܘܕܝܠ ܕܠܫܢܐ: BEREL ܡܠܦܐ ܝܠܗ ܥܠ ܬܠܡܘܕ، ܡܕܪܫ ܘ-Rishonim – ܥܒܪܝܐ ܘܐܪܡܝܐ ܓܕܝܠ̈ܐ – ܘܐܗܘ ܝܠܗ ܫܬܐܣܐ ܕܡܐܢ̈ܐ ܕ-Dicta ܩܐ ܢܘܩܕܐ، ܩܐ ܦܘܣܩܐ ܘܩܐ ܦܘܪܫܐ ܕܟܬܝ̈ܒܬܐ ܡܩܒܠ̈ܢܝܬܐ ܓܘ ܐܢܝ ܟܬܝ̈ܒܬܐ. ܩܐ ܐܝܢܐ ܕܟܒܨܐ "ܐܝܡܢ ܡܠܬܐ ܡܬܚܙܝܐ ܠܩܕܡܝܬܐ" – ܠܟܣܝܩܘܢ ܬܫܥܝܬܝܐ ܕ-Academy (Ma'agarim)؛ ܘܩܐ ܐܝܢܐ ܕܣܢܝܩܐ ܝܠܗ ܠܟܬܝ̈ܒܬܐ ܓܢܗܝܢ، ܒܡܛܝܬܐ ܕܬܘܟܢܝܬܐ – API ܕ-Sefaria.

7. ܛܟܣܐ ܕܥܒܕܐ: ܦܘܫܩܐ ܟܡܝܝܐ ܕܟܬܝܒܬܐ ܥܒܪܝܬܐ ܓܘ ܒܘܨܝܐ ܟܢܫܝܐ

ܗܘܝܐ ܛܘܦܣܢܝܐ: 400 ܦܘܢ̈ܝܐ ܦܬܝ̈ܚܐ ܓܘ ܒܘܨܝܐ ܕܫܘܐ̈ܠܐ (survey)، 30 ܟܬܝ̈ܒܬܐ ܕܡܫܐܠܬ̈ܐ، ܝܢ ܐܠܦܐ posts. ܢܝܫܐ: ܕܡܛܚ ܡܢ ܟܬܝܒܬܐ ܠܡܢܝܢ̈ܐ ܕܡܨܝܐ ܕܡܫܬܕܪܝ ܠܡܚܫܒ̈ܢܐ ܕܣܐܝܛ – ܣܓܝܐܘ̈ܬܐ (frequencies) ܕܓܢ̈ܣܐ (categories)، ܡܢܝ̈ܢܐ ܕܪܓܫܐ ܩܐ ܟܠ ܡܫܬܘܬܦܢܐ (participant)، ܐܣܘܪܐ ܒܝܢܝ ܣܘܟܠܐ ܘܡܫܬܚܠܦ̈ܢܐ ܕܪܩܥܐ. ܫܒܥܐ ܦܣܥ̈ܬܐ، ܘܟܠ ܚܕܐ ܡܢܝܗܝܢ ܟܬܝܒܬܐ ܝܠܗ̇.

ܛܥܝ̈ܘܬܐ ܕܡܬܚܙܝܢ ܣܓܝ (Common mistakes)

ܕܡܢܝܬܘܢ tokens ܘܚܫܒܝܬܘܢ ܕܡܢܐܠܘܟܘܢ ܡ̈ܠܐ. "בבית"، "לבית"، "הבית"، "ביתנו" – ܐܪܒܥܐ ܣܕܪ̈ܐ ܓܘ ܠܘܚܐ ܕܣܓܝܐܘ̈ܬܐ، ܚܕܐ ܡܠܬܐ ܒܫܪܪܐ. ܟܠ ܦܘܫܩܐ ܕܫܒܩ segmentation ܟܬܢܐ ܥܠ ܦܘܠܓܐ (distribution) ܕ-prepositions، ܠܐ ܕܣܘܟܠܐ.

ܕܦܠܚܝܬܘܢ ܒ-stemmer. ܡܐܢ̈ܐ ܓܘܢ̈ܝܐ ܟܡܩܪܒܝ "ܫܩܠܬܐ ܕ-suffixes ܩܐ ܥܒܪܝܐ". ܠܝܬ ܡܕܡ ܐܝܟ ܗܢܐ ܓܘ morphology ܕܫܪܫܐ ܘܡܫܩܠܐ؛ ܦܐܪܐ ܝܠܗ ܣܕܪ̈ܐ ܕܐܬ̈ܘܬܐ ܕܠܐ ܝܢܐ ܡ̈ܠܐ.

ܕܡܫܕܪܝܬܘܢ ܠ-LIWC ܝܢ ܠܠܟܣܝܩܘܢ ܕܪ̈ܓܫܐ ܡܬܪܓܡܐ. ܠܝܬ ܠܟܣܝܩܘܢ LIWC ܥܒܪܝܐ ܪܫܡܝܐ: ܓܙܐ ܕܠܟܣܝܩܘ̈ܢܐ ܕ-LIWC-22 ܟܡܢܐ 18 ܠܫܢ̈ܐ (ܐܝܠܘܠ 2026)، ܘܥܒܪܝܐ ܠܐ ܝܠܗ̇ ܒܝܢܬܝܗܝܢ. ܓܘܕܐ ܝܚܝܕܝܬܐ ܕܬܪܓܡܠܗ̇ ܠܗ ܓܡܝܪܐܝܬ – Shapira ܘܐܚܪ̈ܢܐ (2021) – ܥܒܕܠܗ̇ ܐܗܐ ܥܡ 150 ܐܠܦ̈ܐ ܕܬܘܪ̈ܓܡܐ ܡܩܪ̈ܒܐ ܘܡܩܪܒܘܬ 1,000 ܫܥ̈ܐ ܕܕܝܢܐ، ܘܠܐ ܐܝܬ ܠܗ̇ ܐܦܣܢܘܬܐ ܕܡܦܪܣܐ ܠܦܐܪܐ ܡܛܠ ܕ-LIWC ܬܓܪܝܐ ܝܠܗ (ܦܐܬܐ 56، 62). ܐܦ ܟܡܦܫܩܐ ܩܐ ܡܘܕܝ ܬܘܪܓܡܐ ܡܠܬܐ ܡܠܬܐ ܟܦܐܫ ܡܚܝܐ: "ܬܘܪܓܡܐ ܬܪܝܨܐ ܟܡܛܐ ܠܦܐܪ̈ܐ ܡܢܬܢܝ̈ܐ ܝܢ ܛܥܝ̈ܐ" (ܦܐܬܐ 55) – ܨܘܪ̈ܬܐ ܕܫܡ̈ܐ ܘܕܦ̈ܥܠܐ ܕܠܝܬ ܠܗܘܢ ܡܩܒܠܢܐ ܐܢܓܠܝܙܝܐ، ܡ̈ܠܐ ܦܪ̈ܝܫܐ ܕܟܡܬܚܠܛܝ ܓܘ ܐܦ ܗܝ lemma، ܘܠܟܣܝܩܘܢ "ܐܢܐ" ܕ-LIWC (12 ܡ̈ܠܐ ܒܐܢܓܠܝܙܝܐ) ܕ-morphology ܥܒܪܝܬܐ "ܟܟܠܝܐ ܠܗ ܡܢ ܡܢܝܢܐ" (ܦܐܬܐ 56). ܡܐ ܐܝܬ ܒܕܘܟܬܗ: Hebrew Psychological Lexicons – ܠܟܣܝܩܘ̈ܢܐ ܕܪ̈ܓܫܐ، depression ܘܫܝܢܐ ܕܢܦܫܐ ܕܒܢܝ̈ܐ ܝܢܐ ܩܐ ܥܒܪܝܐ ܡܢ 872 ܟܬܝ̈ܒܬܐ ܕ-psychotherapy ܘܡܫܪܪ̈ܐ ܝܢܐ ܒܝܕ ܬܠܬܐ ܕܝܢ̈ܐ (κ ܕ-Fleiss .54 ܓܘ ܚܘܕܪܐ ܩܕܡܝܐ، .95 ܒܬܪ ܫܘܝܢܐ؛ ܦܐܬܐ 57)، ܡܓܢ ܓܘ GitHub (CC BY-SA؛ ܨܪܪܐ hepsylex). ܩܐ HebEMO ܐܝܬ ܠܦܚܘܬ ܡܢ ܟܠ ܝܘܠܦܢܐ ܥܒܪܝܐ – ܘܗܠܝܐ ܣܢܝܩܐ ܝܠܗ ܠܫܘܪܪܐ (ܦܣܩܐ 7، ܦܣܥܬܐ 6).

ܕܬܟܝܬܘܢ ܥܠ "ܥܘܕܪܢܐ ܩܐ ܥܒܪܝܐ" ܓܘ ܡܐܢܐ ܓܘܢܝܐ. ܓܘ spaCy ܥܒܪܝܐ "none yet" ܝܠܗ̇؛ ܓܘ AntConc ܘܓܘ Voyant ܥܒܪܝܐ ܟܡܬܛܥܢܐ ܐܢܐ ܠܐ ܥܡ segmentation؛ ܓܘ ܡܘܕܝ̈ܠܐ ܕܠܫܢ̈ܐ ܣܓܝ̈ܐܐ ܟܡܬܦܫܩܐ ܒܝܕ ܓܙܐ ܕܡ̈ܠܐ ܐܢܓܠܝܙܝܐ. ܥܘܕܪܢܐ = tokenization، ܐܠܐ ܐܢ ܐܡܝܪܐ ܝܠܗ ܡܕܡ ܐܚܪܢܐ.

ܕܡܪܗܛܝܬܘܢ ܡܘܕܝܠ ܚܕܬܐ ܥܠ ܟܬܝܒܬܐ ܠܐ ܚܕܬܐ. AlephBERT ܥܠ ܦܐܬܐ ܕܓܡܪܐ، HebPipe ܥܠ ܨܚܚܐ ܓܘ ܐܫܥܝܐ – ܡܦܩܢܐ ܦܝܫܐ ܡܬܚܙܝܐ ܬܪܝܨܐ ܘܦܝܫܐ ܛܥܝܐ ܒܝܬܪܘܬܐ. ܙܢܐ ܕܠܫܢܐ ܩܕܡ ܡܐܢܐ (ܦܣܩܐ 1ܕ).

ܕܠܐ ܥܒܕܝܬܘܢ normalization – ܘܫܩܠܝܬܘܢ ܬܪܬܝܢ ܡ̈ܠܐ ܩܐ ܟܠ ܡܠܬܐ. ܢܘܩܕܐ ܕܠܐ ܫܩܝܠܐ ܝܠܗ، ܓܪܫܝܝܡ ܕܠܐ ܚܘܝܕܐ ܝܢܐ، ܟܬܝܒܬܐ ܡܠܝܬܐ ܘܚܣܝܪܬܐ ܚܕܐ ܓܢܒ ܚܕܐ. ܐܬܐ: ܡܠܬܐ ܕܣܓܝ ܡܬܚܙܝܐ ܡܢ ܟܠ ܟܡܬܚܙܝܐ ܒܣܓܝܐܘܬܐ ܒܨܝܪܬܐ ܕܬܗܪܐ.

ܕܝܗܒܝܬܘܢ ܦܣܩܐ ܠܡܘܕܝܠ. ܡܢܝܢܐ ܕ-topics، ܬܚܘܡܐ ܕ-sentiment، ܪܫܝܡܬܐ ܕܡ̈ܠܐ ܕܟܠܝܐ – ܦܣ̈ܩܐ ܕܒܘܨܝܐ ܝܢܐ، ܕܟܬܝ̈ܒܐ ܘܡܥܠ̈ܠܐ ܝܢܐ. ܡܘܕܝܠ ܕܓܒܐ ܠܗ "ܒܝܚܕܝܘܬܗ" ܡܘܕܝܠ ܝܠܗ ܕܓܒܐ ܠܗ ܓܘܒܝܐ ܡܩܕܡܐ (default) ܕܚܕ ܐܚܪܢܐ.

ܪܫܝܡܬܐ ܕܒܘܚܪܢܐ (checklist) – ܩܕܡ ܦܘܫܩܐ ܕܟܬܝܒܬܐ ܥܒܪܝܬܐ

ܡܒܘ̈ܥܐ

Academy of the Hebrew Language. (2017). ܢܡܘ̈ܣܐ ܕܟܬܝܒܬܐ ܡܠܝܬܐ (Kelalei ha-ktiv ha-male). https://hebrew-academy.org.il/topic/hahlatot/missingvocalizationspelling/

Adler, M., & Elhadad, M. (2006). An unsupervised morpheme-based HMM for Hebrew morphological disambiguation. In Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics (pp. 665-672). Association for Computational Linguistics. https://doi.org/10.3115/1220175.1220259

Ahia, O., Kumar, S., Gonen, H., Kasai, J., Mortensen, D. R., Smith, N. A., & Tsvetkov, Y. (2023). Do all languages cost the same? Tokenization in the era of commercial language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (pp. 9904-9923). Association for Computational Linguistics. https://doi.org/10.18653/v1/2023.emnlp-main.614

Bareket, D., & Tsarfaty, R. (2021). Neural modeling for named entities and morphology (NEMO²). Transactions of the Association for Computational Linguistics, 9, 909-928. https://doi.org/10.1162/tacl_a_00404

Chriqui, A., & Yahav, I. (2022). HeBERT and HebEMO: A Hebrew BERT model and a tool for polarity analysis and emotion recognition. INFORMS Journal on Data Science, 1(1), 81-95. https://doi.org/10.1287/ijds.2022.0016

Drucker, J., & Albrezzi, F. (2026). The digital humanities coursebook: Applied concepts and critical approaches (2nd ed.). Routledge. https://doi.org/10.4324/9781003589792

Goldin, G., Howell, N., Ordan, N., Rabinovich, E., & Wintner, S. (2025). The Knesset corpus: An annotated corpus of Hebrew parliamentary proceedings. Language Resources and Evaluation, 59(3), 2973-3004. https://doi.org/10.1007/s10579-025-09833-4

Goregaokar, M., & Leroy, R. (Eds.). (2026). Unicode bidirectional algorithm (Unicode Standard Annex No. 9, Revision 52). Unicode Consortium. https://www.unicode.org/reports/tr9/

Itai, A., & Wintner, S. (2008). Language resources for Hebrew. Language Resources and Evaluation, 42(1), 75-98. https://doi.org/10.1007/s10579-007-9050-8

Kingham, C., & van Peursen, W. (2018). The ETCBC database of the Hebrew Bible. Journal for Semitics, 27(1), Article 2974. https://doi.org/10.25159/1013-8471/2974

McCarthy, J. J. (1981). A prosodic theory of nonconcatenative morphology. Linguistic Inquiry, 12(3), 373-418.

More, A., Seker, A., Basmova, V., & Tsarfaty, R. (2019). Joint transition-based models for morpho-syntactic parsing: Parsing strategies for MRLs and a case study from Modern Hebrew. Transactions of the Association for Computational Linguistics, 7, 33-48. https://doi.org/10.1162/tacl_a_00253

Neguse, T. T., Gabay, H., Reychav, I., McHaney, R., & Jonathan, G. M. (2026). AI-based chatbots and public value creation in digital government services. Transforming Government: People, Process and Policy. Advance online publication. https://doi.org/10.1108/TG-02-2026-0074

Pérez Fernández, M. (1997). An introductory grammar of Rabbinic Hebrew (J. Elwolde, Trans.). Brill. (Original work published 1992)

Ravid, D., & Schiff, R. (2006). Roots and patterns in Hebrew language development: Evidence from written morphological analogies. Reading and Writing, 19(8), 789-818. https://doi.org/10.1007/s11145-006-9004-3

Roorda, D. (2017). The Hebrew Bible as data: Laboratory – sharing – experiences. In J. Odijk & A. van Hessen (Eds.), CLARIN in the Low Countries (pp. 217-229). Ubiquity Press. https://doi.org/10.5334/bbi.18

Roorda, D. (2018). Coding the Hebrew Bible. Research Data Journal for the Humanities and Social Sciences, 3(1), 27-41. https://doi.org/10.1163/24523666-01000011

Rust, P., Pfeiffer, J., Vulić, I., Ruder, S., & Gurevych, I. (2021). How good is your tokenizer? On the monolingual performance of multilingual language models. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) (pp. 3118-3135). Association for Computational Linguistics. https://doi.org/10.18653/v1/2021.acl-long.243

Sáenz-Badillos, A. (1993). A history of the Hebrew language (J. Elwolde, Trans.). Cambridge University Press. (Original work published 1988)

Seker, A., Bandel, E., Bareket, D., Brusilovsky, I., Greenfeld, R., & Tsarfaty, R. (2022). AlephBERT: Language model pre-training and evaluation from sub-word to sentence level. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (pp. 46-56). Association for Computational Linguistics. https://doi.org/10.18653/v1/2022.acl-long.4

Shapira, N., Atzil-Slonim, D., Juravski, D., Baruch, M., Stolowicz-Melman, D., Paz, A., Alfi-Yogev, T., Azoulay, R., Singer, A., Revivo, M., Dahbash, C., Dayan, L., Naim, T., Gez, L., Yanai, B., Maman, A., Nadaf, A., Sarfati, E., Baloum, A., . . . Goldberg, Y. (2021). Hebrew psychological lexicons. In Proceedings of the Seventh Workshop on Computational Linguistics and Clinical Psychology: Improving Access (pp. 55-69). Association for Computational Linguistics. https://doi.org/10.18653/v1/2021.clpsych-1.6

Shmidman, A., Guedalia, J., Shmidman, S., Shmidman, C. S., Handel, E., & Koppel, M. (2022). Introducing BEREL: BERT embeddings for rabbinic-encoded language. arXiv. https://arxiv.org/abs/2208.01875

Shmidman, A., Shmidman, S., Koppel, M., & Goldberg, Y. (2020). Nakdan: Professional Hebrew diacritizer. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: System Demonstrations (pp. 197-203). Association for Computational Linguistics. https://doi.org/10.18653/v1/2020.acl-demos.23

Shmidman, S., Shmidman, A., Cohen, A. D. N., & Koppel, M. (2024). Adapting LLMs to Hebrew: Unveiling DictaLM 2.0 with enhanced vocabulary and instruction capabilities. arXiv. https://arxiv.org/abs/2407.07080

Shmidman, S., Shmidman, A., & Koppel, M. (2023). DictaBERT: A state-of-the-art BERT suite for Modern Hebrew. arXiv. https://arxiv.org/abs/2308.16687

Shmidman, S., Shmidman, A., Koppel, M., & Tsarfaty, R. (2024). MRL parsing without tears: The case of Hebrew. In Findings of the Association for Computational Linguistics: ACL 2024 (pp. 4537-4550). Association for Computational Linguistics. https://doi.org/10.18653/v1/2024.findings-acl.269

Sima'an, K., Itai, A., Winter, Y., Altman, A., & Nativ, N. (2001). Building a tree-bank of Modern Hebrew text. Traitement Automatique des Langues, 42(2), 347-380.

Talstra, E., & Sikkel, C. J. (2000). Genese und Kategorienentwicklung der WIVU-Datenbank, oder: Ein Versuch, dem Computer Hebräisch beizubringen. In C. Hardmeier, W.-D. Syring, J. D. Range, & E. Talstra (Eds.), Ad fontes! Quellen erfassen – lesen – deuten. Was ist Computerphilologie? Ansatzpunkte und Methodologie – Instrumente und Praxis (Applicatio 15, pp. 33-68). VU University Press.

Tsarfaty, R., Bareket, D., Klein, S., & Seker, A. (2020). From SPMRL to NMRL: What did we learn (and unlearn) in a decade of parsing morphologically-rich languages (MRLs)? In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (pp. 7396-7408). Association for Computational Linguistics. https://doi.org/10.18653/v1/2020.acl-main.660

Tsarfaty, R., Sadde, S., Klein, S., & Seker, A. (2019). What's wrong with Hebrew NLP? And how to make it right. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP): System Demonstrations (pp. 259-264). Association for Computational Linguistics. https://doi.org/10.18653/v1/D19-3044

Unicode Consortium. (2026). The Unicode Standard, Version 18.0.0. https://www.unicode.org/versions/Unicode18.0.0/

Whistler, K. (Ed.). (2026). Unicode normalization forms (Unicode Standard Annex No. 15, Revision 58). Unicode Consortium. https://www.unicode.org/reports/tr15/

Wintner, S. (2004). Hebrew computational linguistics: Past and future. Artificial Intelligence Review, 21(2), 113-138. https://doi.org/10.1023/B:AIRE.0000020865.73561.bc

Zeldes, A., Howell, N., Ordan, N., & Ben Moshe, Y. (2022). A second wave of UD Hebrew treebanking and cross-domain parsing. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing (pp. 4331-4344). Association for Computational Linguistics. https://doi.org/10.18653/v1/2022.emnlp-main.292