Elamur Research⌕
☰
home / papers / 1985 / Grammatical Analysis by Computer of the Lancaster-Oslo/Bergen (LOB) Corpus of British English Texts

Grammatical Analysis by Computer of the Lancaster-Oslo/Bergen (LOB) Corpus of British English Texts

Paper · 1985 · Andrew David

Summary

The paper reports on Lancaster's work, with partners in Oslo and Bergen, to annotate the LOB corpus automatically. LOB is a British counterpart to the Brown Corpus: about 1 million words of text printed in 1961. The core problem is that English gives few inflectional cues to word class, and many word forms belong to several classes.

The pipeline has two steps. First, a lookup stage assigns each token its candidate tags, using a list of word-final character strings and an exception wordlist. Second, the right tag is chosen with a matrix of tag-pair transition probabilities, which finds the most probable path through runs of ambiguous words.

This statistical step alone tagged over 93% of tokens correctly. Two additions raised accuracy by roughly 3 more points: retagging a few fixed multiword sequences first, and adjusting weights for some three-tag patterns. Human post-editors fixed the rest, and far fewer corrections were needed than for Brown.

The paper also covers a revised, more mnemonic and hierarchical tag set, an interactive mode, and the start of a hand-parsed treebank. The treebank is meant to supply statistics for probabilistic phrase and clause parsing.

Key points

  • LOB has 500 samples of about 2,000 words each, spread over 15 text categories in informative and imaginative prose. It mirrors the Brown Corpus design so British and American English can be compared.
  • The tagging lexicon has a suffixlist of about 7,000+ strings of 1–5 final characters and a wordlist of about 7,000 entries. The wordlist also holds every form seen more than 50 times. In total, 135 word tags are used.
  • The earlier Greene and Rubin approach used 5,300 ordered context-frame rules and reached 77% success. The Lancaster probabilistic method got over 93%, and about 96% after the refinements.
  • Manual post-editing interventions fell from about 230,000 for Brown to about 35,000 for LOB. Without pre-editing, Knut Hofland estimated success at about 94%, roughly 2 points lower.
  • A treebank of about 35,000 words (1,500 sentences) had been hand-parsed under Sampson's Case-law Manual. The target is about 50,000 words, an even ~5% cross-section of the corpus categories.

Interesting ideas

  • I1. Rare readings get only coarse marks in the lexicon: '@' for under 1 in 10 and '%' for under 1 in 100. These halve or divide the probability by eight, so the rare tag wins only when context strongly supports it.
  • I2. No transition value may fall to zero, so an unseen tag pair is unlikely but never impossible. The system always returns some analysis rather than halting on odd syntax.
  • I3. A small list of fixed multiword sequences (e.g. "according to", "as well as") is retagged before statistical disambiguation, removing predictable errors upstream.
  • I4. Error statistics were used to revise not just the lexicon and matrix but the scoring formula itself. Raw conditional frequency was replaced by a pair-frequency ratio normalized by both tags' frequencies.
  • I5. A hand-drawn treebank sample was built partly to stress-test the annotation rulebook against real sentence variety, and partly to bootstrap parse statistics. Post-editing decisions were also logged as explicit correctness criteria.

For Mova

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

  • I1 (tokens): The en lexicon could carry similar cheap rarity flags on secondary POS readings, giving the tagger an explainable prior instead of opaque per-word counts.
  • I2 (tree): The perceptron parser and gparse could apply a minimum floor score to rare arc or label combinations, keeping output total on eccentric sentences while flagging them as low confidence.
  • I3 (tokens): A curated multiword-expression pass before tagging would keep idioms out of the statistical tagger and make those decisions traceable to one list entry.
  • I4 (learn): Evaluation reports should be allowed to challenge the scoring function, not only the data; for example, compare normalized and conditional association scores when ranking silver-data agreements.
  • I5 (learn): Each LLM-teacher or manual correction round could record the rule it applied, turning repairs into a growing guideline that also seeds self-induced repair rules.

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 UniversityLancaster · United Kingdom36 works · 1,615 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 1Research subjects and languages the work is about.English · 2,379 works

Cite

Cite as Andrew David. 1985. Grammatical Analysis by Computer of the Lancaster-Oslo/Bergen (LOB) Corpus of British English Texts. 23rd Annual Meeting of the Association for Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/981210.981246

@inproceedings{david1985grammatical,
  title = {{Grammatical Analysis by Computer of the Lancaster-Oslo/Bergen (LOB) Corpus of British English Texts}},
  author = {Andrew David},
  booktitle = {23rd Annual Meeting of the Association for Computational Linguistics},
  year = {1985},
  doi = {10.3115/981210.981246},
  url = {https://doi.org/10.3115/981210.981246}
}

References

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

  1. The first phrase of the pruject, completed in September 1983, produced a grammatically annotated version of the corpus giving a tag showing the word class of each word token. Over 93 per cent of the word tags were correctly selected by using a matrix of tag pair probabilities and this figure was upgraded by a further 3 per cent by retagging problematic strings of words prior to disambiguation and by altering the probability weightings for sequences of three tags. The remaining 3 to per cent were corrected by a human post-editor. find
  2. The system was originally designed to run in batch mode over the corpus but we have recently modified procedures to run interactively for sample sentences typed in by a user at a terminal. We are currently extending the word tag set and improving the word tagging procedures to further reduce manual intervention. A similar probabilistic system is being developed for phrase and clause tagging. ~qE STI~JCTURE A~D PURPOSE OF THE LOB CORPUS. The LOB Corpus (Johansson, Leech and Goodluck, 1978), like its American ~/gl~sh counterpart, the Brown Corpus LKucera and Francis, 196a; Hauge and ;Iofland, 1978), is a collection of 500 samples of British ~hglish texts, each containing about 2,000 word tokens. The samples are representations of 15 different ~ext categories: A. Press (Reportage); B. Press (Editorial); C. Press (Reviews); D. Religion; E. i l l s and Hobbies; F. Popular Lore; G. Belles Lettres, Biography, r'[emoirs, etc. ; H. Miscellaneous ; J. find
  3. Learned and Scientific; K. General Fiction; L. Mystery and Detective Fiction; M. Science Fiction; N. Adventure and Western Fiction, Romance and Love Story; R. Humour. There are two main sections, informative prose and imaginative prose, and all the texts contained in the corpus weee printed in a single year (1961). find
  4. The structure of the LOB corpus was designed to resemble that of the Brown corpus as closely as possible so that a systematic comparison of British and American written English could be made. Both corpora contain samples of texts published in the same year (1961) so that comparisons are not distorted by diachronic factors. find
  5. The LOB corpus is used as a database for linguistic research and language description. Historically, different ]inguists have been concerned to a greater or lesser extent with the use of corpus citations, to some degree, at least, because of differences in the perceived view of the descriptive requirements of grammar. Jespersen (1909-A9), Kruisinga and Erades (1911) gave frequent examples of citations from assembled corpora of written texts to illustrate grammatical rules. Work on text corpora is, of course, very much alive toda~v. Storage, retrieval and processing of natural language text is a more efficient and less laborious task with modern computer hardware than it was with hand-written card files but data capture is still a significant problem (Francis, 1980). find
  6. The forthcoming work, A Comprehensive Grammar of the E l i s h Lan~la~e (Quirk, Greenbaum, leech, and ~arr.vik, 1985) contains many citations from both LOB and Brown Corpora. find
  7. A GRAF~ATICALLY A N N O T A OF E CORPUS VERSION Since 1981, research has been directed towards writing programs to grammatically annotate the LOB cor~is. From 1981-83, the research effort produced a version of the corpus with every word token labelled by a grammatical tag showing the word class of each word form. Subsequent research has attempted to build on the techni~les used for automatic word tagging by using the output from the word tagging programs as input to phrase and clause tagging and by using probabilistic methods to provide a constituent analysis of the LOB corpus. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/david-1985-grammatical-analysis-computer-lancaster-oslo/