[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82020-en":3,"doc-seo-82020-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},82020,7971461740886,"Theodore","https://ap-avatar.wpscdn.com/davatar_3d24733baf745e90a7e4bdd5f77d97b2",8,"Research & Report","Holographic Neural PCFG for Unsupervised Parsing","Unsupervised constituency parsing induces latent syntactic tree structures from raw text without annotations. Neural PCFG parameterizations achieve strong results but score grammar rules with high-capacity black-box networks, leaving rule probabilities without an interpretable mathematical form. Holographic Neural PCFG (Hol-PCFG) models PCFG rule scoring as algebraic relations among grammar-symbol embeddings using torus-constrained holographic embeddings, producing closed-form rule probabilities. Hol-PCFG reaches state-of-the-art results across six languages, reduces rule-scoring parameters by 99.94%, improves training stability, and parses Japanese from characters without morphological segmentation.","Holographic Neural PCFG for Unsupervised Parsing  \nRyosuke Yamaki 1 Daichi Mochihashi2 Nobutaka Shimada 1 Tadahiro Taniguchi3 , 1  \n1Ritsumeikan University, Japan 2The Institute of Statistical Mathematics, Japan  \n3 Kyoto University, Japan  \n[yamaki.ryosuke@em.ci.ritsumei.ac.jp](yamaki.ryosuke@em.ci.ritsumei.ac.jp), [daichi@ism.ac.jp](daichi@ism.ac.jp)  \n[shimada@ci.ritsumei.ac.jp](shimada@ci.ritsumei.ac.jp), [taniguchi@i.kyoto-u.ac.jp](taniguchi@i.kyoto-u.ac.jp)  \narXiv :2607 .08063v 1 [ cs .CL] 9 Jul 2026  \nAbstract  \nUnsupervised constituency parsing aims to accurately induce latent tree structures from raw text alone. Recent neural parameterizations of PCFGs achieve strong performance in both supervised and unsupervised parsing, yet rely on high-capacity black-box networks for rule scoring—as exemplified by the Neural PCFG family—leaving rule probabilities without an interpretable mathematical form. In this paper, we propose Holographic Neural PCFG (Hol-PCFG), which recasts PCFG rule scoring as algebraic relation modeling among grammar-symbol embeddings. Hol-PCFG adapts Holographic Embeddings (Nickel et al., 2016), which scores knowledge-graph triples via circular correlation, to the leftchild, right-child, and lexical-emission relations over torus-constrained embeddings, giving every rule probability a closed form that carries the intrinsic structure of grammar rules by construction. HolPCFG achieves state-of-the-art parsing performance in six languages while cutting rule-scoring parameters by 99.94% relative to the baseline model and training more stably. Additionally, we demonstrate that Hol-PCFG can parse Japanese directly from characters without any morphological segmentation, retaining nearly the same morpheme-level performance.  \n1 Introduction  \nUnsupervised constituency parsing is the task of inducing the syntactic tree structure of natural language from raw text alone without any annotations. This task is not merely a benchmark for parsing accuracy, but embodies a fundamental question about language acquisition: whether compact, interpretable discrete grammars can be induced only from word-level distributional information. This question remains relevant even in the era of large language models (LLMs): LLM-based  \nunsupervised parsing methods (Cao et al., 2020 ; Li and Lu, 2023 ; Chen et al., 2024) can induce parse trees by exploiting rich pretrained knowledge, but they do not necessarily learn an explicit generative grammar of reusable nonterminal symbols and rule probabilities. PCFG-based unsupervised parsing therefore remains important for low-resource languages and computational models of language acquisition (Bannard et al., 2009 ; Jin et al., 2021a,b) .  \nThe Neural PCFG (N-PCFG; Kim et al., 2019a) family, the focus of this work, retains an explicit discrete grammar while parameterizing rule probabilities with neural embeddings. N-PCFG, however, treats each child pair of a binary rule as anatomic unit with its own embedding, so the parameter count of the rule scorer grows quadratically with the number of nonterminal and preterminal symbols, restricting early models to only a few dozen such symbols. Subsequent work showed that induction quality improves substantially with more symbols, and scaled grammarsto thousands of nonterminals—TN-PCFG (Yang et al., 2021b) via a low-rank decomposition of the rule tensor, and SN-PCFG (Liu et al., 2023) via a conditional-independence assumption between the left and right children, combined with efficient GPU computation. A complementary recent direction is SemInfo (Chen et al., 2025), which addresses the mismatch between likelihood improvements and parsing accuracy by proposing a new objective function that directly exploits the semantic information content of induced constituents.  \nThe structure of the rule-scoring function itself, however, has received comparatively little attention: the advances above target grammar size and training objectives. In all of these models,","cbCaia1i2JfHDQVW","https://ap.wps.com/l/cbCaia1i2JfHDQVW","pdf",1367849,4,1,13,"English","en",105,"# Abstract\n# Introduction","[{\"question\":\"What problem does unsupervised constituency parsing address?\",\"answer\":\"It induces the syntactic tree structure from raw text alone, without any gold annotations. The goal is to learn latent grammar structures directly from word-level distributional information.\"},{\"question\":\"Why are traditional Neural PCFG rule scorers considered less interpretable?\",\"answer\":\"Many neural approaches use high-capacity MLPs to score PCFG rules, treating scoring as generic function approximation. This yields rule probabilities that lack a closed-form algebraic meaning derived from symbol embeddings.\"},{\"question\":\"What is the core idea of Hol-PCFG?\",\"answer\":\"Hol-PCFG recasts PCFG rule scoring as algebraic relation modeling between grammar-symbol embeddings on a torus. It uses holographic embedding relations (via circular correlation) so every normalized rule probability has a closed form that preserves grammar-rule structure by design.\"}]",1784177627,33,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"holographic-neural-pcfg-for-unsupervised-parsing","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/holographic-neural-pcfg-for-unsupervised-parsing/82020/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-29","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does unsupervised constituency parsing address?","Question",{"text":75,"@type":76},"It induces the syntactic tree structure from raw text alone, without any gold annotations. The goal is to learn latent grammar structures directly from word-level distributional information.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"Why are traditional Neural PCFG rule scorers considered less interpretable?",{"text":80,"@type":76},"Many neural approaches use high-capacity MLPs to score PCFG rules, treating scoring as generic function approximation. This yields rule probabilities that lack a closed-form algebraic meaning derived from symbol embeddings.",{"name":82,"@type":73,"acceptedAnswer":83},"What is the core idea of Hol-PCFG?",{"text":84,"@type":76},"Hol-PCFG recasts PCFG rule scoring as algebraic relation modeling between grammar-symbol embeddings on a torus. It uses holographic embedding relations (via circular correlation) so every normalized rule probability has a closed form that preserves grammar-rule structure by design.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]