---
title: "Learning unification-based grammars using the Spoken English"
description: "The paper addresses a gap in grammar induction: most systems of the time learned performance… 24 references, cited by 1 paper here."
authors: ["Miles Osborne", "Derek Bridge"]
year: 1994
venue: "arXiv:cmp-lg/9406040 · ICGI-94 Colloquium"
source: "https://arxiv.org/abs/cmp-lg/9406040"
doi: "10.48550/arxiv.cmp-lg/9406040"
arxiv: "cmp-lg/9406040"
licence: "arXiv — no licence stated (read only)"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/osborne-1994-learning-unification-based-grammars-using/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/osborne-1994-learning-unification-based-grammars-using/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [1994](https://md.research.elamur.ai/free/en/papers/1994/) / Learning unification-based grammars using the Spoken English Corpus

# Learning unification-based grammars using the Spoken English Corpus

**Authors:** [Miles Osborne](https://md.research.elamur.ai/free/en/authors/miles-osborne/) and Derek Bridge · **Year:** 1994 · **Venue:** [arXiv:cmp-lg/9406040 · ICGI-94 Colloquium](https://md.research.elamur.ai/free/en/venues/lecture-notes-in-computer-science/) ([Lecture notes in computer science 1994](https://md.research.elamur.ai/free/en/venues/lecture-notes-in-computer-science/1994/)) · **Publisher:** [Springer Science+Business Media](https://md.research.elamur.ai/free/en/publishers/springer-science-business-media/) · **Licence:** arXiv — no licence stated (read only)

**Links:** [Publisher's page](https://arxiv.org/abs/cmp-lg/9406040) · [PDF](https://arxiv.org/pdf/cmp-lg/9406040) · [DOI 10.48550/arxiv.cmp-lg/9406040](https://doi.org/10.48550/arxiv.cmp-lg/9406040) · [arXiv cmp-lg/9406040](https://arxiv.org/abs/cmp-lg/9406040) · [OpenAlex](https://openalex.org/W1532084191)

**Citations:** 6 citations (OpenAlex, 2026-10-14) · 1 in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper addresses a gap in grammar induction: most systems of the time learned performance grammars, while the authors wanted competence grammars that also give linguistically sensible parse structures. Their learner starts from a hand-written unification-based grammar. When a sentence cannot be parsed, it fills the gaps with maximally general "super rules" and then prunes the proposed rule instances. Two kinds of evidence drive the pruning. One is a linguistic model: GPSG LP rules, typed λ-calculus semantic types, a Head Feature Convention and X-bar syntax. The other is a data-driven score built from mother-daughter pair frequencies. The system extends the GDE and was tested on the tagged and parsed Spoken English Corpus (SEC). Learning tripled coverage on held-out sentences. Using both pruning sources together produced more plausible parses than either source alone, although every learnt grammar still scored below the hand-built starting grammar. The work suggests that theory-driven and corpus-driven constraints can work together to extend formal grammars for corpus parsing.

## Key points
- **Resource:** The SEC contains about 50,000 words of broadcast monologue. It comes already tagged and parsed, so no separate lexicon or gold evaluation set had to be built.
- **Implementation and setup:**
  - The system adds about 3300 lines of Common Lisp to the GDE.
  - The starting grammar had 97 rules over the CLAWS2 tagset.
  - The model used 4 LP rules, 32 semantic types and an HFC. X-bar syntax was always on.
- **Data splits:** The SEC sentences were divided into three disjoint sets: pretrain (fewer than 20 sentences), train (60) and test (60).
- **Results by configuration (size / coverage / plausibility):**

  | Configuration | Rules | Test coverage | Plausibility |
  |---|---|---|---|
  | No learning | 97 | 26.7% | 0.103 |
  | Data-driven only | 129 | 75.0% | 0.086 |
  | Model-based only | 128 | 65.0% | 0.095 |
  | Combined | 129 | 75.0% | 0.098 |

- **Search limits and caveats:**
  - To keep search tractable, learning stopped after 1 parse or 3000 edges, and only the binary super rule was used.
  - The low plausibility scores come partly from a mismatch: the grammar produces deep, fine-grained trees, while the SEC's hand parses are shallow, coarse and of uneven quality.

## Interesting ideas
- When parsing fails, the system proposes completely unconstrained rules that join any partial constituents, then lets independent filters remove implausible instances, so generation and validation stay separate modules.
- Compositional semantic types serve as a veto on syntax: a rule is rejected if the meanings of its daughters cannot combine.
- Subtrees are scored recursively from parent–child label pair frequencies, using a geometric mean so that nodes with more children are not unfairly penalised. Unseen pairs get a small nonzero score so that every tree can still be ranked.
- An unusually high count of chart edges is treated as a sign that the input may be ungrammatical, not just as a resource limit.
- Learnt-grammar parses are evaluated only on sentences that need at least one learnt rule. A separate yardstick set, parsed by the original grammar, serves as the baseline, and coverage gains are reported alongside structural quality.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (learn): Self-induced repair rules could follow this pattern. A failed parse would trigger generic attachment proposals, which would then be pruned by explicit, explainable constraints before being kept as new rules.
- **I2** (frames): Candidate frames or role assignments could be discarded when the argument types (for example, VerbNet selectional restrictions) cannot fill the frame. This gives a traceable reason for each rejection.
- **I3** (tree): Head–dependent relation-pair statistics from treebanks could offer a cheap, inspectable plausibility score for rescoring candidate UD trees from gparse or the perceptron parser.
- **I4** (tree): Search-effort counters could flag sentences that are likely malformed or out of distribution, so they are routed to repair or excluded from silver data.
- **I5** (learn): Evaluation could check every new rule or model on the inputs it newly covers and compare against a baseline control set. This shows whether coverage was bought at the cost of tree quality.

*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 York](https://md.research.elamur.ai/free/en/universities/university-of-york/) — University, York, United Kingdom · 17 works · 6,570 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (1): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,263 works

## Cite

Cite as: Miles Osborne and Derek Bridge. 1994. Learning unification-based grammars using the Spoken English Corpus. arXiv:cmp-lg/9406040 · ICGI-94 Colloquium.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{osborne1994learning,
  title = {{Learning unification-based grammars using the Spoken English Corpus}},
  author = {Miles Osborne and Derek Bridge},
  year = {1994},
  doi = {10.48550/arxiv.cmp-lg/9406040},
  eprint = {cmp-lg/9406040},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.cmp-lg/9406040}
}
```

- [← previous Computational Analyses of Arabic Morphology](https://md.research.elamur.ai/free/en/papers/kiraz-1994-computational-analyses-arabic-morphology/)
- [next → Evaluation Metrics for Knowledge-Based Machine Translation](https://md.research.elamur.ai/free/en/papers/mitamura-1994-evaluation-metrics-knowledge-based-machine/)
- [↑ back to papers of 1994](https://md.research.elamur.ai/free/en/papers/1994/)

## References

24 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. J. K. Baker. Trainable grammars for speech recognition. In D. H. Klatt and J. J. Wolf, editors, Speech Communication Papers for the 97 th Meeting of the Acoustical Society of America, pages 547–550. 1979. [find](https://scholar.google.com/scholar?q=J%2E+K%2E+Baker%2E+Trainable+grammars+for+speech+recognition%2E+In+D%2E+H%2E+Klatt+and+J%2E+J%2E+Wolf%2C+editors%2C+Speech+Communication+Papers+for+the+97+th+Meeting+of+the+Acoustical+Society+of+America%2C+pages+547%E2%80%93550%2E+1)
2. Robert C. Berwick. The acquisition of syntactic knowledge. MIT Press, 1985. [↗](https://doi.org/10.7551/mitpress/1074.001.0001)
3. Ezra Black, Roger Garside, and Geoffrey Leech, editors. Statistically driven computer grammars of English the IBM-Lancaster approach. Rodopi, 1993. [find](https://scholar.google.com/scholar?q=Ezra+Black%2C+Roger+Garside%2C+and+Geoffrey+Leech%2C+editors%2E+Statistically+driven+computer+grammars+of+English+the+IBM%2DLancaster+approach%2E+Rodopi%2C+1993%2E)
4. Eric Brill, David Magerman, Mitchell Marcus, and Beatrice Santorini. Deducing Linguistic Structure from the Statistics of Large Corpora. In AAAI-92 Workshop Program: Statistically-Based NLP Techniques, San Jose, California, 1992. [find](https://scholar.google.com/scholar?q=Eric+Brill%2C+David+Magerman%2C+Mitchell+Marcus%2C+and+Beatrice+Santorini%2E+Deducing+Linguistic+Structure+from+the+Statistics+of+Large+Corpora%2E+In+AAAI%2D92+Workshop+Program%3A+Statistically%2DBased+NLP+Techniques)
5. Claudia Casadio. Semantic Categories and the Development of Categorial Grammars. In Richard T. Oehrle, editor, Categorial Grammars and Natural Language Structures, pages 95–123. D. Reidel, 1988. [↗](https://doi.org/10.1007/978-94-015-6878-4_5)
6. John Carroll, Claire Grover, Ted Briscoe, and Bran Boguraev. A Development Environment for Large Natural Language Grammars. Technical report number 127, University of Cambridge Computer Laboratory, 1988. [find](https://scholar.google.com/scholar?q=John+Carroll%2C+Claire+Grover%2C+Ted+Briscoe%2C+and+Bran+Boguraev%2E+A+Development+Environment+for+Large+Natural+Language+Grammars%2E+Technical+report+number+127%2C+University+of+Cambridge+Computer+Laboratory%2C+19)
7. Noam Chomsky. Lectures on Government and Binding. Dordrecht: Foris, 1981. [find](https://scholar.google.com/scholar?q=Noam+Chomsky%2E+Lectures+on+Government+and+Binding%2E+Dordrecht%3A+Foris%2C+1981%2E)
8. K. Church and R. Patil. Coping with syntactic ambiguity or how to put the block in the box on the table. Computational Linguistics, 8:139–49, 1982. [find](https://scholar.google.com/scholar?q=K%2E+Church+and+R%2E+Patil%2E+Coping+with+syntactic+ambiguity+or+how+to+put+the+block+in+the+box+on+the+table%2E+Computational+Linguistics%2C+8%3A139%E2%80%9349%2C+1982%2E)
9. Robert de Beaugrande and Wolfgang Dressler. Introduction to Text Linguistics. Longman, 1981. [↗](https://doi.org/10.4324/9781315835839)
10. D.R. Dowty, R.E. Wall, and S. Peters. Introduction to Montague Semantics. D. Reidel Publishing Company, 1981. [↗](https://doi.org/10.1007/978-94-009-9065-4)
11. G. Gadzar, E. Klein, G.K. Pullum, and I.A. Sag. Generalized Phrase Structure Grammar. Harvard University Press, 1985. [find](https://scholar.google.com/scholar?q=G%2E+Gadzar%2C+E%2E+Klein%2C+G%2EK%2E+Pullum%2C+and+I%2EA%2E+Sag%2E+Generalized+Phrase+Structure+Grammar%2E+Harvard+University+Press%2C+1985%2E)
12. R. Garside, G. Leech, and G. Sampson, editors. The Computational Analysis of English: A Corpus-based Approach. Longman, 1987. [find](https://scholar.google.com/scholar?q=R%2E+Garside%2C+G%2E+Leech%2C+and+G%2E+Sampson%2C+editors%2E+The+Computational+Analysis+of+English%3A+A+Corpus%2Dbased+Approach%2E+Longman%2C+1987%2E)
13. M. A. K. Halliday and Ruqaiya Hasan. Coherence in English. Longman, 1976. [find](https://scholar.google.com/scholar?q=M%2E+A%2E+K%2E+Halliday+and+Ruqaiya+Hasan%2E+Coherence+in+English%2E+Longman%2C+1976%2E)
14. Ray S. Jackendoff. X-Bar Syntax: A Study of Phrase Structure. The M.I.T Press, 1977. [find](https://scholar.google.com/scholar?q=Ray+S%2E+Jackendoff%2E+X%2DBar+Syntax%3A+A+Study+of+Phrase+Structure%2E+The+M%2EI%2ET+Press%2C+1977%2E)
15. S. Johansson, G. Leech, and H. Goodluck. Manual of Information to Accompany the Lancaster-Oslo/Bergen Corpus of British English, for Use with Digital Computers. Technical report, Department of English, University of Oslo, 1978. [find](https://scholar.google.com/scholar?q=S%2E+Johansson%2C+G%2E+Leech%2C+and+H%2E+Goodluck%2E+Manual+of+Information+to+Accompany+the+Lancaster%2DOslo%2FBergen+Corpus+of+British+English%2C+for+Use+with+Digital+Computers%2E+Technical+report%2C+Department+of+English)
16. Fanny Leech. An approach to probabilistic parsing. MPhil Dissertation, 1987. University of Lancaster. [find](https://scholar.google.com/scholar?q=Fanny+Leech%2E+An+approach+to+probabilistic+parsing%2E+MPhil+Dissertation%2C+1987%2E+University+of+Lancaster%2E)
17. Geoffrey Leech and Roger Garside. Running a grammar factory: The production of syntactically analysed corpora or “treebanks”. In Stig Johansson and Anna-Brita Stenström, editors, English Computer Corpora: Selected Papers and Research Guide. Mouten de Gruyter, 1991. [↗](https://doi.org/10.1515/9783110865967.15)
18. K. Lari and S. J. Young. The estimation of stochastic context-free grammars using the Inside-Outside Algorithm. Computer Speech and Language, 4:35–56, 1990. [↗](https://doi.org/10.1016/0885-2308(90)90022-x)
19. D. Magerman and M. Marcus. Pearl: a probabilistic chart parser. In Proceedings of the 2 nd International Workshop on Parsing Technologies, Cancun, Mexico, pages 193–199, 1991. [find](https://scholar.google.com/scholar?q=D%2E+Magerman+and+M%2E+Marcus%2E+Pearl%3A+a+probabilistic+chart+parser%2E+In+Proceedings+of+the+2+nd+International+Workshop+on+Parsing+Technologies%2C+Cancun%2C+Mexico%2C+pages+193%E2%80%93199%2C+1991%2E)
20. David Magerman and Carl Weir. Efficiency, Robustness and Accuracy in Picky Chart Parsing. In Proceedings of the 30 th ACL, University of Delaware, Newark, Delaware, pages 40–47, 1992. [find](https://scholar.google.com/scholar?q=David+Magerman+and+Carl+Weir%2E+Efficiency%2C+Robustness+and+Accuracy+in+Picky+Chart+Parsing%2E+In+Proceedings+of+the+30+th+ACL%2C+University+of+Delaware%2C+Newark%2C+Delaware%2C+pages+40%E2%80%9347%2C+1992%2E)
21. G. Numberg. The linguistics of punctuation. Center for the Study of Language and Information, 1990. [find](https://scholar.google.com/scholar?q=G%2E+Numberg%2E+The+linguistics+of+punctuation%2E+Center+for+the+Study+of+Language+and+Information%2C+1990%2E)
22. Miles Osborne and Derek Bridge. Inductive and deductive grammar learning: dealing with incomplete theories. In Grammatical Inference Colloquim, Essex University, 1993. [find](https://scholar.google.com/scholar?q=Miles+Osborne+and+Derek+Bridge%2E+Inductive+and+deductive+grammar+learning%3A+dealing+with+incomplete+theories%2E+In+Grammatical+Inference+Colloquim%2C+Essex+University%2C+1993%2E)
23. Miles Osborne and Derek Bridge. Learning unification-based grammars and the treatment of undergeneration. In Workshop on Machine Learning Techniques and Text Analysis, Vienna, Austria, 1993. [find](https://scholar.google.com/scholar?q=Miles+Osborne+and+Derek+Bridge%2E+Learning+unification%2Dbased+grammars+and+the+treatment+of+undergeneration%2E+In+Workshop+on+Machine+Learning+Techniques+and+Text+Analysis%2C+Vienna%2C+Austria%2C+1993%2E)
24. Kurt Vanlehn and William Ball. A Version Space Approach to Learning Context-free Grammars. Machine Learning, 2.1:39–74, 1987. [↗](https://doi.org/10.1007/bf00058754)

*Updated 1 November 2026*