--- title: "Constraint Grammar as a Framework for Parsing Running Text" description: "He proposes Constraint Grammar (CG), a declarative and language-independent formalism. 5 references, cited by 76 papers here." authors: ["Fred Karlsson"] year: 1990 venue: "COLING 1990 Volume 3: Papers presented to the 13th International Conference on Computational Linguistics" source: "https://aclanthology.org/C90-3030/" licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)" licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/karlsson-1990-constraint-grammar-as-framework-parsing/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/karlsson-1990-constraint-grammar-as-framework-parsing/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [1990](https://md.research.elamur.ai/free/en/papers/1990/) / Constraint Grammar as a Framework for Parsing Running Text # Constraint Grammar as a Framework for Parsing Running Text **Authors:** Fred Karlsson · **Year:** 1990 · **Venue:** [COLING 1990 Volume 3: Papers presented to the 13th International Conference on Computational Linguistics](https://md.research.elamur.ai/free/en/venues/coling/) ([COLING 1990](https://md.research.elamur.ai/free/en/venues/coling/1990/)) · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/) **Links:** [Publisher's page](https://aclanthology.org/C90-3030/) · [PDF](https://aclanthology.org/C90-3030.pdf) · [OpenAlex](https://openalex.org/W2046592963) **Citations:** 206 citations (OpenAlex, 2026-10-14) · [76](https://md.research.elamur.ai/free/en/papers/karlsson-1990-constraint-grammar-as-framework-parsing/cited-by/) in our library, of the 141,840 papers whose bibliographies we read ## Summary Karlsson addresses the problem of parsing unrestricted running text. He argues that grammars imported from theories like GB, GPSG or LFG are poorly suited to the task, mainly because they handle ambiguity badly. He proposes Constraint Grammar (CG), a declarative and language-independent formalism. Input tokens are already analysed by two-level morphological analysers (Koskenniemi's model), and every word starts with all its possible readings, including candidate syntactic function labels drawn from the lexicon and from morphosyntactic mapping. Corpus-derived constraints then remove readings that do not fit the context. One mechanism handles morphological disambiguation, clause boundary detection and surface syntactic function assignment. The output is a single flat, annotated string, and any ambiguity that cannot be resolved stays visible in it. The parser, CGP, is a Common Lisp interpreter. For English, morphological disambiguation comes close to the 95–99% reported for probabilistic taggers. Syntactic labelling leaves slightly over 90% of words with a unique label. The work matters because it shows that a fairly small set of linguist-written, morphology-driven constraints can compete with statistical tagging while also covering shallow syntax. ## Key points - The pipeline has these stages: - preprocessing (including multi-word expressions such as *in spite of*) - morphological analysis - local disambiguation - morphosyntactic mapping - "syntax proper", which covers context-dependent disambiguation, clause boundaries and syntactic functions. - Two-level morphological analysers exist for English (41,000 lexicon entries), Finnish (37,000) and Swedish (42,000). - Local disambiguation was tested on a Swedish text of about 840,000 tokens: - Morphology assigned roughly 1.5 million readings, about 1.8 per word-form. - Local disambiguation removed over 100,000 of them. - Cohorts with 6 readings dropped from 3,830 to 312. - A constraint has four parts: domain, operator, target and context conditions. - There are three operators: `=!!`, `=!` and `=0`. - Context positions can be absolute (e.g. -1, 3) or unbounded with `*`, and relative positions can be anchored with `L`. - For English syntax, about 220 mapping statements and 25 syntactic constraints give just over 90% uniquely labelled words. The author expects around 500 disambiguation constraints to be enough in the end. ## Interesting ideas - **Remove readings instead of building structures.** All alternatives are present from the start, and rules only delete them. This guarantees exactly one output per sentence and avoids an explosion of parse trees. The approach transfers well to any annotation pipeline that must stay robust on messy real-world text. - **Prefer the analysis with the fewest compound boundaries.** For languages with productive compounding, the reading with the fewest compound boundaries is kept. Likewise, non-derived readings win over derived ones. This cheap, context-free heuristic is reported as nearly perfect, and similar rules could prune analyser output for other highly inflected or compounding languages. - **Use one formalism for morphology and syntax.** Syntactic functions are treated as labels of the same kind as morphological features. The same constraint machinery can then resolve both kinds of ambiguity, which keeps the system simple and lets each layer help the other. - **Exploit uniqueness within the clause.** Some functions, such as subject or finite main verb, should occur only once per clause. Checking this repeatedly lets the parser safely discard competing labels. Structural cardinality constraints like this are a cheap pruning signal in many tagging tasks. - **Make risk an explicit choice.** Grammar writers decide deliberately whether a rule may fire on an ambiguous context, and truly undecidable cases, such as PP attachment, are left unresolved. This trades coverage against precision in a transparent way, which is useful when downstream users need reliable labels. *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* ## Institutions - [University of Helsinki](https://md.research.elamur.ai/free/en/universities/university-of-helsinki/) — University, Helsinki, Finland · 33 works · 4,705 citations 📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works ## Cite Cite as: Fred Karlsson. 1990. Constraint Grammar as a Framework for Parsing Running Text. COLING 1990 Volume 3: Papers presented to the 13th International Conference on Computational Linguistics. The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{karlsson1990constraint, title = {{Constraint Grammar as a Framework for Parsing Running Text}}, author = {Fred Karlsson}, booktitle = {COLING 1990 Volume 3: Papers presented to the 13th International Conference on Computational Linguistics}, year = {1990}, doi = {10.3115/991146.991176}, url = {https://doi.org/10.3115/991146.991176} } ``` - [← previous A Spelling Correction Program Based on a Noisy Channel Model](https://md.research.elamur.ai/free/en/papers/kernighan-1990-spelling-correction-program-based-noisy/) - [next → Semantic-Head-Driven Generation](https://md.research.elamur.ai/free/en/papers/shieber-1990-semantic-head-driven-generation/) - [↑ back to papers of 1990](https://md.research.elamur.ai/free/en/papers/1990/) ## References 5 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. Church, K. 1988. "A Stochastic Parts Program and Noun Phrase Parser for Running Text." SecondCon-ference on Applied Natural Language Processing, Proceedings of the Conference, ACL 1988, pp.136-143. [find](https://scholar.google.com/scholar?q=Church%2C+K%2E+1988%2E+%22A+Stochastic+Parts+Program+and+Noun+Phrase+Parser+for+Running+Text%2E%22+SecondCon%2Dference+on+Applied+Natural+Language+Processing%2C+Proceedings+of+the+Conference%2C+ACL+1988%2C+pp%2E136%2D143%2E) 2. Garside, Roger, Leech, Geoffrey, and Sampson, Geoffrey 1987 (eds.), 7he Computational Analysis of English. [find](https://scholar.google.com/scholar?q=Garside%2C+Roger%2C+Leech%2C+Geoffrey%2C+and+Sampson%2C+Geoffrey+1987+%28eds%2E%29%2C+7he+Computational+Analysis+of+English%2E) 3. Longman, London and New York. Karlsson, Fred 1989. "Parsing and Constraint Grammar". Manuscript, Research Unit for Computational Linguistics, University of Helsinki. [find](https://scholar.google.com/scholar?q=Longman%2C+London+and+New+York%2E+Karlsson%2C+Fred+1989%2E+%22Parsing+and+Constraint+Grammar%22%2E+Manuscript%2C+Research+Unit+for+Computational+Linguistics%2C+University+of+Helsinki%2E) 4. Koskenniemi, Kimmo 1983. Two-Level Morphology. A General Computational Model for Word-Form Recognition and Production. Department of General Linguistics, University of Helsinki, Publications No. 13. [find](https://scholar.google.com/scholar?q=Koskenniemi%2C+Kimmo+1983%2E+Two%2DLevel+Morphology%2E+A+General+Computational+Model+for+Word%2DForm+Recognition+and+Production%2E+Department+of+General+Linguistics%2C+University+of+Helsinki%2C+Publications+No%2E+13%2E) 5. Acknowledgements This research was supported by the Academy of Finland in 1985-89, and by the Technology Development Centre of Finland (TEKES) in 1989-90. Part of it belongs to the ESPRIT II project SIMPR (2083). I am indebted to Kimmo Koskenniemi for help in the field of morphological analysis, and… [find](https://scholar.google.com/scholar?q=Acknowledgements+This+research+was+supported+by+the+Academy+of+Finland+in+1985%2D89%2C+and+by+the+Technology+Development+Centre+of+Finland+%28TEKES%29+in+1989%2D90%2E+Part+of+it+belongs+to+the+ESPRIT+II+project+S) *Updated 24 October 2026*