Mova0.5 pre-release⌕
☰
home / docs / papers / authors / Reut Tsarfaty

Reut Tsarfaty

31 paper cards in Mova's library · 2006–2026

Reut Tsarfaty wrote 31 of the papers Mova reads, in 2006–2026, most of them on Treebanks, Dependency parsing, Named entity recognition; often with Amit Seker, Dan Bareket, Avi Shmidman.

OpenAlex · ORCID · h-index 24 · 2886 citations · 217 works — openalex, 2026-10-18

The profile is the library's match of the name; two people with one name are not always told apart.

Elsewhere: Google Scholar · Semantic Scholar · dblp · ACL Anthology · ORCID

Topics: Treebanks 6 · Dependency parsing 4 · Named entity recognition 3 · Part-of-speech tagging 3 · Coreference resolution 1 · Universal Dependencies 1

Ideas Mova took (6)

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

Papers

  1. Beyond Word Boundaries: A Hebrew Coreference Benchmark and an Evaluation Protocol for Morphologically Complex Text (2026)
    The paper builds KibutzR, a Modern Hebrew coreference dataset on the UD Hebrew treebank. The 6,151 sentences are regrouped into 351 documents, and any document that appears in more than one partition is moved wholly into training. Mentions are marked at word, morpheme and multiword level: a possessive suffix or…
  2. A Truly Joint Neural Architecture for Segmentation and Parsing (2024)
    The paper argues that Hebrew segmentation and dependency parsing should be solved jointly, as pre-neural parsers did.
  3. HeSum: a Novel Dataset for Abstractive Text Summarization in Hebrew (2024)
    HeSum is a Modern Hebrew summarisation set of 10,000 article–summary pairs (8,000 train, 1,000 validation, 1,000 test) from three independent news sites, where the summary is the editor's extended subheading.
  4. 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.
  5. HeSum: a Novel Dataset for Abstractive Text Summarization in Hebrew (2024)
    The paper builds HeSum, 10,000 Modern Hebrew news articles paired with the editor-written extended subheadings that serve as summaries, taken from three independent news sites and split 8,000 / 1,000 / 1,000.
  6. Multilingual Sequence-to-Sequence Models for Hebrew NLP (2023)
    The authors argue that Hebrew encoder-only models (AlephBERT, AlephBERT-Gimmel, HeBERT) cannot emit morphemes or labels that are not in the input, so they need extra segment-then-tag components.
  7. 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.
  8. HeGeL: A Novel Dataset for Geo-Location from Hebrew Text (2023)
    The task is to turn a free-text place description into map coordinates. The authors crowdsourced 5,649 literal Hebrew place descriptions for 167 places in Tel Aviv, Haifa and Jerusalem. Writers described a place they know without naming it or any landmark within 100 m; a second participant then marked it on…
  9. HeQ: a Large and Diverse Hebrew Reading Comprehension Benchmark (2023)
    HeQ is an extractive question-answering benchmark in the SQuAD style for Modern Hebrew: 30,147 question–answer pairs from Hebrew Wikipedia and from the Geektime tech news site.
  10. Explicit Morphological Knowledge Improves Pre-training of Language Models for Hebrew (2023)
    The question is whether a BERT-style Hebrew model learns better when its tokenizer is told about Hebrew morphology before pre-training, instead of a purely statistical WordPiece split.
  11. The Truth, The Whole Truth, and Nothing but the Truth: A New Benchmark Dataset for Hebrew Text Credibility Assessment (2023)
    The paper introduces HeTrue, 1,021 statements by Israeli politicians and public figures, collected and graded by the journalists of an IFCN fact-checking outlet.
  12. AlephBERT: Language Model Pre-training and Evaluation from Sub-Word to Sentence Level (2022) 1 idea kept
    Seker et al. pre-train AlephBERT, a BERT-style model for Modern Hebrew, on about 1.9 billion words: the deduplicated Hebrew part of OSCAR, all of Hebrew Wikipedia, and Hebrew tweets, with a 52K wordpiece vocabulary. The base model trained for 10 epochs and took 8 days on eight V100 GPUs. The paper's main point is…
  13. 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.
  14. AlephBERT: A Hebrew Large Pre-Trained Language Model to Start-off your Hebrew NLP Application With (2021)
    The authors build AlephBERT, a BERT-base style language model for Modern Hebrew, pre-trained on about 1.9 billion words: deduplicated OSCAR, a large set of Hebrew tweets (72% of sentences), and Hebrew Wikipedia.
  15. Neural Modeling for Named Entities and Morphology (NEMO2) (2021)
    In Hebrew, named-entity boundaries often fall inside a space-delimited token, because a preposition or the definite article is attached to the name.
  16. The Possible, the Plausible, and the Desirable: Event-Based Modality Detection for Language Processing (2021)
    The paper defines event-based modality detection: find every modal trigger in a sentence, of any part of speech (should, plan, prefers, probably, necessary, able), give it a sense, and link it to the head of the event it modifies.
  17. Evaluating Models' Local Decision Boundaries via Contrast Sets (2020) measured
    Standard test sets come from the same distribution as the training data, so a dataset with systematic gaps or annotation artifacts lets models score well with shallow heuristics.
  18. A Pointer Network Architecture for Joint Morphological Segmentation and Tagging (2020) 3 ideas kept
    Morphological disambiguation (MD) picks one decomposition of each token into morphemes. The authors feed a symbolic lattice of all analyses per token (HebLex for Hebrew, TRMorph2 for Turkish) into a neural pointer network. The lattice is flattened into a sequence of analyses, each analysis is embedded by averaging…
  19. pyBART: Evidence-based Syntactic Transformations for IE (2020)
    pyBART is an open-source Python library that converts English UD trees into a representation (BART) that makes event structure and many content-word relations explicit, going beyond Enhanced UD.
  20. 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…
  21. Joint Transition-Based Models for Morpho-Syntactic Parsing: Parsing Strategies for MRLs and a Case Study from Modern Hebrew (2019)
    In Hebrew, morphological analysis and disambiguation (MA&D) normally comes before parsing, yet the syntax would help choose the morphology, and the reverse.
  22. What's Wrong with Hebrew NLP? And How to Make it Right (2019)
    The paper presents ONLP (YAP), an open system for Modern Hebrew that does segmentation, POS tagging, lemmas, features and labelled dependencies in one package.
  23. Representations and Architectures in Neural Sentiment Analysis for Morphologically Rich Languages: A Case Study from Modern Hebrew (2018)
    The paper studies how two input choices change neural sentiment analysis for Modern Hebrew: token-based or morpheme-based input (morphemes from the YAP parser), and string-based or character-based vocabulary.
  24. The Hebrew Universal Dependency Treebank: Past, Present and Future (2018)
    The Hebrew Treebank (HTB) grew from 500 newspaper sentences (2001) to 6221 sentences, through a chain of automatic and semi-automatic conversions: phrase structure, then Goldberg's functional-head dependencies, then Unified-Stanford, then UD v1 and UD v2.
  25. Universal Morpho-Syntactic Parsing and the Contribution of Lexica: Analyzing the ONLP Lab Submission to the CoNLL 2018 Shared Task (2018)
    The ONLP Lab submission used UDPipe for segmentation and morphological disambiguation, then yap's arc-eager dependency parser (a Zhang–Nivre variant with morphological-signature features).
  26. Universal Joint Morph-Syntactic Processing: The Open University of Israel's Submission to The CoNLL 2017 Shared Task (2017)
    The OpenU-NLP-Lab system for the CoNLL 2017 UD shared task is a transition-based, non-neural joint morphological disambiguator and arc-standard dependency parser (built in yap).
  27. Universal Dependencies v1: A Multilingual Treebank Collection (2016)
    The paper is the reference for the first UD release. It sets out the design principles: dependency-based and lexicalist annotation, the recoverability principle (every syntactic word must map back to the raw text), content words as the backbone of the tree with function words as dependents, and a closed inventory of…
  28. Evaluating Dependency Parsing: Robust and Heuristics-Free Cross-Annotation Evaluation (2011)
    Attachment scores compare parsers only within one treebank scheme. Two parsers trained on different conversions of the Penn Treebank can differ mainly because of annotation choices, such as the head of a prepositional phrase or of a copula. The authors propose TEDEVAL: each dependency tree is mapped by deterministic…
  29. Modeling Morphosyntactic Agreement in Constituency-Based Parsing of Modern Hebrew (2010)
    The question is whether agreement features (gender, number, person, definiteness, accusativity) help a statistical parser of Hebrew, and how they should be represented.
  30. Word-Based or Morpheme-Based? Annotation Strategies for Modern Hebrew Clitics (2008)
    The Modern Hebrew Treebank splits pronominal clitics on prepositions, accusative markers and possessive markers into separate leaves: one orthographic word becomes two or more nodes (morpheme-based, MB).
  31. Integrated Morphological and Syntactic Disambiguation for Modern Hebrew (2006)
    Hebrew morphology (prefixed particles, clitics, a fused word that may span several phrases) makes word-level parsing hard.

Co-authors (76)

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