---
title: "A Probabilistic Approach to Grammatical Analysis of Written"
description: "The paper describes how the Lancaster team (UCREL) built a grammatically annotated version of the LOB Corpus of… Cited by 2 papers here."
authors: ["Andrew David Beale"]
year: 1985
venue: "Second Conference of the European Chapter of the Association for Computational Linguistics"
source: "https://aclanthology.org/E85-1023/"
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/beale-1985-probabilistic-approach-grammatical-analysis-written/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/beale-1985-probabilistic-approach-grammatical-analysis-written/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [1985](https://md.research.elamur.ai/free/en/papers/1985/) / A Probabilistic Approach to Grammatical Analysis of Written English by Computer.

# A Probabilistic Approach to Grammatical Analysis of Written English by Computer.

**Authors:** [Andrew David Beale](https://md.research.elamur.ai/free/en/authors/andrew-david-beale/) · **Year:** 1985 · **Venue:** [Second Conference of the European Chapter of the Association for Computational Linguistics](https://md.research.elamur.ai/free/en/venues/eacl/) ([EACL 1985](https://md.research.elamur.ai/free/en/venues/eacl/1985/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **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/E85-1023/) · [PDF](https://aclanthology.org/E85-1023.pdf) · [OpenAlex](https://openalex.org/W2046946758)

**Citations:** 3 citations (OpenAlex, 2026-10-14) · [2](https://md.research.elamur.ai/free/en/papers/beale-1985-probabilistic-approach-grammatical-analysis-written/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper describes how the Lancaster team (UCREL) built a grammatically annotated version of the LOB Corpus of written British English. This was the first step toward software that can analyse unrestricted English text. Word tagging worked in two phases. First, each word received candidate tags from a lexicon of full word forms or, failing that, from a list of word endings. Then the most likely candidate in context was chosen using tag-to-tag transition probabilities taken from the tagged Brown corpus, with a few heuristics added. The tagger got over 96% of tags right, and human post-editors fixed the rest. The paper then extends the same propose-then-disambiguate approach to constituent parsing. Pairs of word tags propose partial structures called T-tags, and the most likely combination is chosen using statistics on which constituents dominate which. These statistics come from a small hand-built treebank following a written rulebook. The work matters because it showed that shallow statistical methods with local context could annotate a million-word corpus at scale.

## Key points
- **LOB Corpus:** 500 samples of about 2,000 words each, published in 1961, spread over 15 text categories and matching the design of the Brown Corpus. It holds just over 1 million tokens.
- **Word tagger:**
  - Written in PASCAL between 1981 and 1983.
  - Candidate tags come from a full-form list or a "suffixlist" of 1–5 final characters.
  - Choices are made with a one-step transition matrix (Marshall, 1983).
  - Accuracy rose from about 93% to over 96% after two fixes: pre-retagging problem sequences and reweighting selected "tag triples".
- **Hypertag set:** each phrase or clause label is one capital letter for the main category, optionally followed by lowercase letters for subcategories (e.g. `Na`, `Vzb`). `Y` acts as a wildcard, and `S` is the root.
- **Treebank:**
  - About 1,500 sentences were parsed by hand by G. Sampson, following the Case Law Manual and documented in the Tree Notes.
  - Each token is stored as four fields: a reference number, the word, its tag, and a hypertag/bracket field.
- **Annotation policy:**
  - Form wins over function, so "in fact" is labelled a prepositional phrase.
  - No paraphrase links are recorded, and deleted elements are mostly not restored.
  - Unary branching is mostly avoided.

## Interesting ideas
- When a word is missing from the lexicon, guess its tags from its last few characters (1–5), so rare or new words still get plausible candidates; this is cheap and robust.
- Give every token a set of candidate analyses and then choose the best path through the whole sequence, so that runs of adjacent ambiguous words are resolved together rather than one at a time.
- Let pairs of adjacent tags propose small fragments of structure (open, close or continue a constituent), and let a global step pick a consistent set; local rules can over-propose because selection prunes later.
- Document annotation decisions as a growing book of precedents, updated whenever a hard corpus case appears, and propagate rule changes back into the existing trees.
- Prefer labelling by form rather than function in the tree, and leave meaning relations such as paraphrase to other layers.

## For Mova

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

- **I1** (tokens): A suffix fallback in crate en could give candidate POS and morphology for out-of-vocabulary words before the contextual tagger decides.
- **I2** (tokens): The tagger could keep scored tag lattices and decode them jointly, which also preserves alternatives for later stages to inspect when explaining a choice.
- **I3** (tree): A rule-table proposer of attachment options could act as an explainable candidate generator or a sanity witness alongside the perceptron parser and gparse.
- **I4** (learn): A versioned "case law" of dialect decisions for silver data and repair rules would keep witness parsers and converted treebanks consistent and make disagreements traceable.
- **I5** (frames): Keeping the UD tree purely formal and pushing functional interpretation into prag makes each layer simpler and its errors easier to attribute.

*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

- [Lancaster University](https://md.research.elamur.ai/free/en/universities/lancaster-university/) — University, Lancaster, United Kingdom · 36 works · 1,615 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (1): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,379 works

## Cite

Cite as: Andrew David Beale. 1985. A Probabilistic Approach to Grammatical Analysis of Written English by Computer. Second Conference of the European Chapter of the Association for Computational Linguistics.

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

```bibtex
@inproceedings{beale1985probabilistic,
  title = {{A Probabilistic Approach to Grammatical Analysis of Written English by Computer.}},
  author = {Andrew David Beale},
  booktitle = {Second Conference of the European Chapter of the Association for Computational Linguistics},
  year = {1985},
  doi = {10.3115/976931.976954},
  url = {https://doi.org/10.3115/976931.976954}
}
```

- [← previous Grammatical Analysis by Computer of the Lancaster-Oslo/Bergen (LOB) Corpus of British English Texts](https://md.research.elamur.ai/free/en/papers/david-1985-grammatical-analysis-computer-lancaster-oslo/)
- [next → An English Generator for a Case-Labelled Dependency Representation](https://md.research.elamur.ai/free/en/papers/tait-1985-english-generator-case-labelled-dependency/)
- [↑ back to papers of 1985](https://md.research.elamur.ai/free/en/papers/1985/)

*Updated 31 October 2026*