Avi Shmidman

Ideas Mova took (2)

From the ideas ledger: an idea tried, kept or queued, or measured with no gain, and what it gave.

• measured, no gain — Constrain prefix segmentation by a table of legal proclitic sets keyed on the word's initial letters. (DictaBERT: A State-of-the-Art BERT Suite for Modern Hebrew · legal prefix-chain table as a hard filter (chain:N, learned chains): split right dev 97.38 / 94.21 unchanged, cv +-0.01 (pointed, unpointed); the learned…)
• kept — Constrain by a wordlist: in unpointed mode, restrict each token's candidates to forms in Mova's lexicon, then score them. (NLP in the DH pipeline: Transfer-learning to a Chronolect · host-lexicon check on unseen words (hostlex:b): hard filter hurts (names cut into known hosts: pointed dev 97.38 -> 95.77); soft b=0.3 kept for pointed (with…)

Papers

1. Dicta-LM 3.0: Advancing The Frontier of Hebrew Sovereign LLMs (2026)
   Dicta-LM 3.0 is an open-weight family of Hebrew-and-English LLMs at 24B, 12B and 1.7B parameters. Each was made by continued pre-training of an existing base model (Mistral-Small-3.1, Nemotron Nano V2, Qwen3-1.7B) on about 100B Hebrew tokens mixed with 30B English tokens, then post-trained with SFT, DPO and GRPO for…
2. Human-AI Annotation Error Auditing for Hebrew Diacritization with Frontier LLMs (2026)
   The authors audit the labels of the EACL 2023 Hebrew Homograph Challenge Set (about 150,000 sentences, niqqud on homographs in Modern Hebrew) for annotation errors.
3. Modeling the Dalet Clitic in Historical Hebrew Texts: A New Prefix-Segmented BERT Model and Stylistic Analysis (2026)
   The Aramaic proclitic dalet in historical Hebrew can mark a subordinate clause or a possessive preposition, and the two uses look the same on the page; no annotated data existed.
4. A New Hebrew Universal Dependency Treebank: The First Treebank of Post-Rabbinic Historical Hebrew (2025)
   The authors release the first UD treebank of post-Rabbinic historical Hebrew: about 11,800 words from six works (Igeret orhot olam, Sefer ha-hinukh, Rashi's Pentateuch commentary, Shulhan arukh, Sefer Maharil and a miscellany), written between the tenth and nineteenth centuries.
5. Automatic Text Segmentation of Ancient and Historic Hebrew (2025)
   Unpunctuated Rabbinic and historic Hebrew and Aramaic texts run to hundreds of words without a full stop, and most neural tools need input cut into chunks.
6. Irony Detection in Hebrew Documents: A Novel Dataset and an Evaluation of Neural Classification Methods (2025)
   Hebrew has no capital letters, so a single word in quotation marks can mean three different things: a name (naming, e.g.
7. NeoDictaBERT: Pushing the Frontier of BERT models for Hebrew (2025)
   The authors move the NeoBERT encoder design (pre-norm RMSNorm, RoPE, SwiGLU, 28 layers at width 768, 4,096-token context) to Hebrew.
8. Restoring Missing Spaces in Scraped Hebrew Social Media (2025)
   Scraped web text often runs two words together, and in unvocalised Hebrew one conflated string can usually be split in several valid ways, so a dictionary lookup does not settle it.
9. MRL Parsing Without Tears: The Case of Hebrew (2024)
   Parsers for morphologically rich languages either run a pipeline (segment, tag, parse), which passes errors forward, or a joint lattice over all segmentations, which is slow and needs a lexicon.
10. MsBERT: A New Model for the Reconstruction of Lacunae in Hebrew Manuscripts (2024)
    The authors pretrain a BERT-base model, MsBERT, on Hebrew manuscript transcriptions. The paper argues that existing Hebrew BERTs, including BEREL trained on printed rabbinic texts, break on manuscript orthography: extra matres lectionis, acronyms written with apostrophes, truncated words, and the proclitic לש. The…
11. Adapting LLMs to Hebrew: Unveiling DictaLM 2.0 with Enhanced Vocabulary and Instruction Capabilities (2024)
    The paper adapts Mistral-7B to Hebrew by continued pre-training on about 200B tokens of half Hebrew and half English.
12. Computational Methods for the Analysis of Complementizer Variability in Language and Literature: The Case of Hebrew "she-" and "ki" (2024)
    Modern Hebrew has two complementizers, še- and ki, that are usually described as interchangeable with ki as the more formal register.
13. OtoBERT: Identifying Suffixed Verbal Forms in Modern Hebrew Literature (2024)
    Modern Hebrew literature sometimes fuses a verb with its direct-object pronoun into one word. The fused form is often identical to a common verb with another meaning, so readers are misled until the sentence ends. Annotators found 222 suffixed verbs in 135K words of literature (17.76 per 10K) against 7 in 185K words…
14. Do Pretrained Contextual Language Models Distinguish between Hebrew Homograph Analyses? (2023)
    Unpointed Hebrew is full of homographs whose analyses differ in segmentation (the prefix or suffix is or is not a word), in morphosyntax, or only in sense.
15. DictaBERT: A State-of-the-Art BERT Suite for Modern Hebrew (2023)
    The paper presents DictaBERT, a BERT-architecture model for Modern Hebrew, pre-trained with masked language modelling only (next-sentence prediction removed) on about 3.8B tokens: 70% cleaned HeDC4 web text and 30% news, blogs, subtitles and novels.
16. Introducing DictaLM -- A Large Generative Language Model for Modern Hebrew (2023)
    The authors release DictaLM, a 7B-parameter generative model for Modern Hebrew, trained on about 7.5 billion tokens.
17. Introducing BEREL: BERT Embeddings for Rabbinic-Encoded Language (2022)
    A short paper releasing a BERT-base model trained on about 220 million words of Rabbinic Hebrew (the Sefaria and Dicta libraries).
18. Large Pre-Trained Models with Extra-Large Vocabularies: A Contrastive Analysis of Hebrew BERT Models and a New One to Outperform Them All (2022)
    The authors train AlephBERTGimmel, a Modern Hebrew BERT with a 128K word-piece vocabulary (earlier models had 2.4K–52K), on the same ~2 billion words as AlephBERT.
19. NLP in the DH pipeline: Transfer-learning to a Chronolect (2021) 1 idea kept
    The authors build a diacritizer (vowel marks) for Emergent Modern Hebrew, the Hebrew of the early twentieth century, using a hand-vocalised goldset of 144,446 words from the Jerusalem Corpus of Emergent Modern Hebrew (129,453 words of literature, 14,993 of ephemera).
20. Nakdan: Professional Hebrew Diacritizer (2020)
    Nakdan adds vowel points to undiacritized Hebrew, for modern, Rabbinic and poetic texts, through a web interface that lets an editor choose among the alternatives.
21. A Novel Challenge Set for Hebrew Morphological Disambiguation and Diacritics Restoration (2020)
    Hebrew homographs are often unbalanced: one analysis is 30 to 129 times more frequent than the other. A tagger can then score well on the majority reading and fail on the minority one, which is the reading that matters in rare cases. The authors build a challenge set of 21 homographs (12 unbalanced, 9 balanced) with…
22. Identification of Parallel Passages Across a Large Hebrew/Aramaic Corpus (2016)
    A fast method for finding all approximate parallel passages (20+ words) in a large Hebrew/Aramaic corpus, where spelling (plene vs defective, final letters, א/י), particles and synonyms vary.

Co-authors (27)

• Moshe Koppel ×13
• Shaltiel Shmidman ×12
• Cheyn Shmuel Shmidman ×4
• Joshua Guedalia ×4
• Reut Tsarfaty ×4
• Hillel Gershuni ×3
• Amir DN Cohen ×2
• Aynat Rubinstein ×2
• Dan Bareket ×2
• Rachel Tal ×2
• Amir David Nissan Cohen
• Amit Seker
• Avishay Gerczuk
• Benjamin Resnick
• Elda Weizman
• Eli Handel
• Elisha Rosensweig
• Elisheva Brauner
• Ely Porat
• Ephraim Meiri
• Eylon Gueta
• Nachum Dershowitz
• Ometz Shmidman
• Orly Albeck
• Shlomit Fuchs
• Yitzchak Lindenbaum
• Yoav Goldberg

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/authors/avi-shmidman/